agent-bot/takt
1
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

workflow 変更

Browse Source
This commit is contained in:
nrslib 2026-01-30 01:50:25 +09:00
parent e67d2648d8
commit f7181fc00c
18 changed files with 1372 additions and 2150 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
resources/global/en/agents/default/ai-antipattern-reviewer.md
View File

@ -126,7 +126,54 @@ AI is confidently wrong—code that looks plausible but doesn't work, solutions
2. Check whether each fallback has a legitimate reason
3. REJECT if even one unjustified fallback exists
### 7. Decision Traceability Review
### 7. Unused Code Detection
**AI tends to generate unnecessary code "for future extensibility", "for symmetry", or "just in case". Delete code that is not called anywhere at present.**
| Judgment | Criteria |
|----------|----------|
| **REJECT** | Public function/method not called from anywhere |
| **REJECT** | Setter/getter created "for symmetry" but never used |
| **REJECT** | Interface or option prepared for future extension |
| **REJECT** | Exported but grep finds no usage |
| OK | Implicitly called by framework (lifecycle hooks, etc.) |
| OK | Intentionally published as public package API |
**Verification approach:**
1. Verify with grep that no references exist to changed/deleted code
2. Verify that public module (index files, etc.) export lists match actual implementations
3. Check that old code corresponding to newly added code has been removed
### 8. Unnecessary Backward Compatibility Code Detection
**AI tends to leave unnecessary code "for backward compatibility." Don't overlook this.**
Code that should be deleted:
| Pattern | Example | Judgment |
|---------|---------|----------|
| deprecated + unused | `@deprecated` annotation with no callers | **Delete immediately** |
| Both new and old API exist | New function exists but old function remains | **Delete old** |
| Migrated wrappers | Created for compatibility but migration complete | **Delete** |
| Comments saying "delete later" | `// TODO: remove after migration` left unattended | **Delete now** |
| Excessive proxy/adapter usage | Complexity added only for backward compatibility | **Replace with simple** |
Code that should be kept:
| Pattern | Example | Judgment |
|---------|---------|----------|
| Externally published API | npm package exports | Consider carefully |
| Config file compatibility | Can read old format configs | Maintain until major version |
| During data migration | DB schema migration in progress | Maintain until migration complete |
**Decision criteria:**
1. **Are there any usage sites?** → Verify with grep/search. Delete if none
2. **Is it externally published?** → If internal only, can delete immediately
3. **Is migration complete?** → If complete, delete
**Be suspicious when AI says "for backward compatibility."** Verify if it's really needed.
### 9. Decision Traceability Review
**Verify that Coder's decision log is reasonable.**

106
resources/global/en/agents/default/architecture-reviewer.md
View File

@ -191,73 +191,6 @@ if (status === 'interrupted') {
}
```
**Fallback Value Overuse Detection Criteria:**
Fallback values (`??`, `||`, default arguments) silently swallow "value is missing" cases. They hide what should be errors.
| Judgment | Criteria |
|----------|----------|
| **REJECT** | Fallback hides a bug where a missing value indicates data inconsistency |
| **REJECT** | Uses meaningless values like `'unknown'`, `'default'`, `''`, `0` as cover |
| **REJECT** | All call sites rely on fallback — no one passes the actual value |
| OK | Defensive default for external input (user input, API responses) |
| OK | Reasonable initial value for optional configuration |
```typescript
// ❌ REJECT - Fallback hiding a bug
const userName = user.name ?? 'unknown'; // Missing name is data inconsistency
const stepName = step?.name ?? 'default'; // Missing step is a caller bug
// ❌ REJECT - Option that all call sites omit
function runStep(step: Step, options?: { maxRetries?: number }) {
const retries = options?.maxRetries ?? 3; // No call site passes options
}
// ✅ OK - Optional user setting with reasonable default
const logLevel = config.logLevel ?? 'info'; // Default if not in config file
const language = userPreference.lang ?? 'en'; // Default if not set
// ✅ OK - Defensive default for external API
const displayName = apiResponse.nickname ?? apiResponse.email; // Fallback if no nickname
```
**Unused Code Detection Criteria:**
AI tends to generate unnecessary code "for future extensibility", "for symmetry", or "just in case". Delete code that is not called anywhere at present.
| Judgment | Criteria |
|----------|----------|
| **REJECT** | Public function/method not called from anywhere |
| **REJECT** | Setter/getter created "for symmetry" but never used |
| **REJECT** | Interface or option prepared for future extension |
| **REJECT** | Exported but grep finds no usage |
| OK | Implicitly called by framework (lifecycle hooks, etc.) |
| OK | Intentionally published as public package API |
```typescript
// ❌ REJECT - Setter "for symmetry" (only get is used)
class WorkflowState {
private _status: Status;
getStatus(): Status { return this._status; }
setStatus(s: Status) { this._status = s; } // No one calls this
}
// ❌ REJECT - Options for "future extension"
interface EngineOptions {
maxIterations: number;
enableParallel?: boolean; // Not implemented. Not referenced anywhere
pluginHooks?: PluginHook[]; // Not implemented. No plugin system exists
}
// ❌ REJECT - Exported but unused
export function formatStepName(name: string): string { ... } // grep result: 0 hits
// ✅ OK - Called by framework
class MyComponent extends React.Component {
componentDidMount() { ... } // Called by React
}
```
**Direct State Mutation Detection Criteria:**
Directly mutating objects or arrays makes changes hard to track and causes unexpected side effects. Always use spread operators or immutable operations to return new objects.
@ -379,36 +312,7 @@ function createUser(data: UserData) {
}
```
### 7. Unnecessary Backward Compatibility Code Detection
**AI tends to leave unnecessary code "for backward compatibility." Don't overlook this.**
Code that should be deleted:
| Pattern | Example | Judgment |
|---------|---------|----------|
| deprecated + unused | `@deprecated` annotation with no callers | **Delete immediately** |
| Both new and old API exist | New function exists but old function remains | **Delete old** |
| Migrated wrappers | Created for compatibility but migration complete | **Delete** |
| Comments saying "delete later" | `// TODO: remove after migration` left unattended | **Delete now** |
| Excessive proxy/adapter usage | Complexity added only for backward compatibility | **Replace with simple** |
Code that should be kept:
| Pattern | Example | Judgment |
|---------|---------|----------|
| Externally published API | npm package exports | Consider carefully |
| Config file compatibility | Can read old format configs | Maintain until major version |
| During data migration | DB schema migration in progress | Maintain until migration complete |
**Decision criteria:**
1. **Are there any usage sites?** → Verify with grep/search. Delete if none
2. **Is it externally published?** → If internal only, can delete immediately
3. **Is migration complete?** → If complete, delete
**Be suspicious when AI says "for backward compatibility."** Verify if it's really needed.
### 8. Workaround Detection
### 7. Workaround Detection
**Don't overlook compromises made to "just make it work."**
@ -423,7 +327,7 @@ Code that should be kept:
**Always point these out.** Temporary fixes become permanent.
### 9. Quality Attributes
### 8. Quality Attributes
| Attribute | Review Point |
|-----------|--------------|
@ -431,7 +335,7 @@ Code that should be kept:
| Maintainability | Easy to modify and fix |
| Observability | Logging and monitoring enabled |
### 10. Big Picture
### 9. Big Picture
**Caution**: Don't get lost in minor "clean code" nitpicks.
@ -442,7 +346,7 @@ Verify:
- Does it align with business requirements
- Is naming consistent with the domain
### 11. Change Scope Assessment
### 10. Change Scope Assessment
**Check change scope and include in report (non-blocking).**
@ -461,7 +365,7 @@ Verify:
**Include as suggestions (non-blocking):**
- If splittable, present splitting proposal
### 12. Circular Review Detection
### 11. Circular Review Detection
When review count is provided (e.g., "Review count: 3rd"), adjust judgment accordingly.

386
resources/global/en/workflows/default.yaml
View File

@ -1,26 +1,36 @@
# Default TAKT Workflow
# Plan -> Coder -> Architect Review -> AI Review -> Security Review -> Supervisor Approval
# Plan -> Coder -> AI Review -> Architect Review -> Security Review -> Supervisor Approval
#
# Template Variables:
# Boilerplate sections (Workflow Context, User Request, Previous Response,
# Additional User Inputs, Instructions heading) are auto-injected by buildInstruction().
# Only step-specific content belongs in instruction_template.
#
# Template Variables (available in instruction_template):
# {iteration} - Workflow-wide turn count (total steps executed across all agents)
# {max_iterations} - Maximum iterations allowed for the workflow
# {step_iteration} - Per-step iteration count (how many times THIS step has been executed)
# {task} - Original user request
# {previous_response} - Output from the previous step
# {previous_response} - Output from the previous step (only when pass_previous_response: true)
# {git_diff} - Current uncommitted changes (git diff)
# {user_inputs} - Accumulated user inputs during workflow
# {report_dir} - Report directory name (e.g., "20250126-143052-task-summary")
#
# Step-level Fields:
# report: - Report file(s) for the step (auto-injected as Report File/Files in Workflow Context)
# Single: report: 00-plan.md
# Multiple: report:
# - Scope: 01-coder-scope.md
# - Decisions: 02-coder-decisions.md
name: default
description: Standard development workflow with planning and specialized reviews
max_iterations: 20
max_iterations: 30
initial_step: plan
steps:
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -42,20 +52,9 @@ steps:
- {Question 2}
pass_previous_response: true
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: plan (Task Analysis)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (when returned from implement)
{previous_response}
## Instructions
Analyze the task and create an implementation plan.
**Note:** If returned from implement step (Previous Response exists),
@ -94,6 +93,9 @@ steps:
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -106,26 +108,10 @@ steps:
permission_mode: acceptEdits
rules:
- condition: Implementation complete
next: review
next: ai_review
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
Follow the plan from the plan step and implement.
Refer to the plan report (00-plan.md) and proceed with implementation.
@ -163,120 +149,9 @@ steps:
- **Reason**: {Why this option was chosen}
```
- name: review
agent: ~/.takt/agents/default/architecture-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
rules:
- condition: No issues found
next: ai_review
- condition: Minor improvements needed
next: improve
- condition: Structural fix required
next: fix
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: review (Architecture Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-architect-review.md
## Original User Request (Initial request from workflow start)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Focus on **architecture and design** review. Do NOT review AI-specific issues (that's the next step).
Review the changes and provide feedback.
**Report output:** Output to the `Report File` specified above.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# Architecture Review
## Result: APPROVE / IMPROVE / REJECT
## Summary
{1-2 sentences summarizing result}
## Reviewed Perspectives
- [x] Structure & Design
- [x] Code Quality
- [x] Change Scope
## Issues (if REJECT)
| # | Location | Issue | Fix |
|---|----------|-------|-----|
| 1 | `src/file.ts:42` | Issue description | Fix method |
## Improvement Suggestions (optional, non-blocking)
- {Future improvement suggestions}
```
**Cognitive load reduction rules:**
- APPROVE + no issues -> Summary only (5 lines or less)
- APPROVE + minor suggestions -> Summary + suggestions (15 lines or less)
- REJECT -> Issues in table format (30 lines or less)
- name: improve
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
permission_mode: acceptEdits
rules:
- condition: Improvements complete
next: review
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: improve
## Architect Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request from workflow start - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Address the Architect's improvement suggestions.
These are minor improvements, not major design issues.
Make improvements such as:
- Naming improvements
- Small refactoring
- Adding/fixing comments
- Code organization
pass_previous_response: true
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
@ -286,26 +161,15 @@ steps:
- WebFetch
rules:
- condition: No AI-specific issues
next: security_review
next: review
- condition: AI-specific issues found
next: ai_fix
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: ai_review (AI-Generated Code Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-ai-review.md
## Original User Request (Initial request from workflow start)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the code for AI-specific issues:
- Assumption validation
- Plausible but wrong patterns
@ -361,21 +225,9 @@ steps:
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: ai_fix
## AI Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request from workflow start - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Address the AI Reviewer's feedback.
Focus on:
- Correcting incorrect assumptions
@ -384,8 +236,125 @@ steps:
- Removing scope creep
pass_previous_response: true
- name: review
agent: ~/.takt/agents/default/architecture-reviewer.md
report: 04-architect-review.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
rules:
- condition: No issues found
next: security_review
- condition: Minor improvements needed
next: improve
- condition: Structural fix required
next: fix
instruction_template: |
## Git Diff
```diff
{git_diff}
```
Focus on **architecture and design** review. Do NOT review AI-specific issues (that's the ai_review step).
Review the changes and provide feedback.
**Report output:** Output to the `Report File` specified above.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# Architecture Review
## Result: APPROVE / IMPROVE / REJECT
## Summary
{1-2 sentences summarizing result}
## Reviewed Perspectives
- [x] Structure & Design
- [x] Code Quality
- [x] Change Scope
## Issues (if REJECT)
| # | Location | Issue | Fix |
|---|----------|-------|-----|
| 1 | `src/file.ts:42` | Issue description | Fix method |
## Improvement Suggestions (optional, non-blocking)
- {Future improvement suggestions}
```
**Cognitive load reduction rules:**
- APPROVE + no issues -> Summary only (5 lines or less)
- APPROVE + minor suggestions -> Summary + suggestions (15 lines or less)
- REJECT -> Issues in table format (30 lines or less)
- name: improve
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
permission_mode: acceptEdits
rules:
- condition: Improvements complete
next: review
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Architect Feedback (This is the latest instruction - prioritize this)
{previous_response}
**Important**: Address the Architect's improvement suggestions.
These are minor improvements, not major design issues.
Make improvements such as:
- Naming improvements
- Small refactoring
- Adding/fixing comments
- Code organization
pass_previous_response: true
- name: fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
permission_mode: acceptEdits
rules:
- condition: Architect feedback addressed
next: review
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Architect Feedback (This is the latest instruction - prioritize this)
{previous_response}
**Important**: Address the Architect's feedback.
The "Original User Request" is reference information, not the latest instruction.
Review the session conversation history and fix the issues raised by the Architect.
pass_previous_response: true
- name: security_review
agent: ~/.takt/agents/default/security-reviewer.md
report: 05-security-review.md
allowed_tools:
- Read
- Glob
@ -399,22 +368,11 @@ steps:
- condition: Vulnerabilities require fix
next: security_fix
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: security_review
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/05-security-review.md
## Original User Request (Initial request from workflow start)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Perform security review on the changes. Check for vulnerabilities including:
- Injection attacks (SQL, Command, XSS)
- Authentication/Authorization issues
@ -473,65 +431,18 @@ steps:
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: security_fix
## Security Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request from workflow start - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the vulnerabilities identified in the security review.
Security issues require highest priority.
pass_previous_response: true
- name: fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
permission_mode: acceptEdits
rules:
- condition: Architect feedback addressed
next: review
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix
## Architect Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request from workflow start - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Address the Architect's feedback.
The "Original User Request" is reference information, not the latest instruction.
Review the session conversation history and fix the issues raised by the Architect.
pass_previous_response: true
- name: supervise
agent: ~/.takt/agents/default/supervisor.md
report:
- Validation: 06-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -546,24 +457,11 @@ steps:
- condition: Requirements unmet, tests failing, build errors
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: supervise (final verification)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/06-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Run tests, verify the build, and perform final approval.
**Workflow Overall Review:**

383
resources/global/en/workflows/expert-cqrs.yaml
View File

@ -1,14 +1,14 @@
# Expert Review Workflow
# Expert CQRS Review Workflow
# Review workflow with CQRS+ES, Frontend, Security, and QA experts
#
# Flow:
# plan -> implement -> cqrs_es_review -> frontend_review -> ai_review -> security_review -> qa_review -> supervise -> COMPLETE
# plan -> implement -> ai_review -> cqrs_es_review -> frontend_review -> security_review -> qa_review -> supervise -> COMPLETE
# ↓ ↓ ↓ ↓ ↓ ↓
# fix_cqrs_es fix_frontend ai_fix fix_security fix_qa fix_supervisor
# ai_fix fix_cqrs_es fix_frontend fix_security fix_qa fix_supervisor
#
# Fix destination is determined by Coder based on change impact:
# - fix_security: MINOR→security_review, MAJOR→cqrs_es_review
# - fix_qa: MINOR→qa_review, SECURITY→security_review, MAJOR→cqrs_es_review
# - fix_security: MINOR->security_review, MAJOR->cqrs_es_review
# - fix_qa: MINOR->qa_review, SECURITY->security_review, MAJOR->cqrs_es_review
#
# Template Variables:
# {iteration} - Workflow-wide turn count (total steps executed across all agents)
@ -23,7 +23,7 @@
name: expert-cqrs
description: CQRS+ES, Frontend, Security, QA Expert Review
max_iterations: 20
max_iterations: 30
initial_step: plan
@ -33,6 +33,7 @@ steps:
# ===========================================
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -42,20 +43,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: plan (Task Analysis)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (when returned from implement)
{previous_response}
## Instructions
Analyze the task and create an implementation plan.
**Note:** If returned from implement step (Previous Response exists),
@ -103,6 +93,9 @@ steps:
# ===========================================
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -113,22 +106,6 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
Follow the plan from the plan step and implement.
Refer to the plan report (00-plan.md) and proceed with implementation.
@ -167,15 +144,105 @@ steps:
```
rules:
- condition: Implementation is complete
next: cqrs_es_review
next: ai_review
- condition: Cannot proceed with implementation
next: plan
# ===========================================
# Phase 2: CQRS+ES Review
# Phase 2: AI Review
# ===========================================
- name: cqrs_es_review
agent: ~/.takt/agents/expert-cqrs/cqrs-es-reviewer.md
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Git Diff
```diff
{git_diff}
```
Review the code for AI-specific issues:
- Assumption validation
- Plausible but wrong patterns
- Context fit with existing codebase
- Scope creep detection
**Report output:** Output to the `Report File` specified above.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# AI-Generated Code Review
## Result: APPROVE / REJECT
## Summary
{One sentence summarizing result}
## Verified Items
| Aspect | Result | Notes |
|--------|--------|-------|
| Assumption validity | ✅ | - |
| API/Library existence | ✅ | - |
| Context fit | ✅ | - |
| Scope | ✅ | - |
## Issues (if REJECT)
| # | Category | Location | Issue |
|---|----------|----------|-------|
| 1 | Hallucinated API | `src/file.ts:23` | Non-existent method |
```
**Cognitive load reduction rules:**
- No issues → Summary 1 line + check table only (10 lines or less)
- Issues found → + Issues in table format (25 lines or less)
rules:
- condition: No AI-specific issues found
next: cqrs_es_review
- condition: AI-specific issues detected
next: ai_fix
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
instruction_template: |
## AI Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
**Important**: Address the AI Reviewer's feedback.
Focus on:
- Correcting incorrect assumptions
- Fixing plausible-but-wrong implementations
- Aligning with existing codebase patterns
- Removing scope creep
pass_previous_response: true
rules:
- condition: AI Reviewer's issues have been fixed
next: ai_review
- condition: Unable to proceed with fixes
next: plan
# ===========================================
# Phase 3: CQRS+ES Review
# ===========================================
- name: cqrs_es_review
agent: ~/.takt/agents/expert-cqrs/cqrs-es-reviewer.md
report: 04-cqrs-es-review.md
allowed_tools:
- Read
- Glob
@ -184,22 +251,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: cqrs_es_review (CQRS+ES Expert Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-cqrs-es-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the changes above from the CQRS (Command Query Responsibility Segregation)
and Event Sourcing perspective.
@ -258,21 +314,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_cqrs_es
## CQRS+ES Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the CQRS+ES expert.
Areas of concern:
@ -289,10 +333,11 @@ steps:
next: plan
# ===========================================
# Phase 3: Frontend Review
# Phase 4: Frontend Review
# ===========================================
- name: frontend_review
agent: ~/.takt/agents/expert/frontend-reviewer.md
report: 05-frontend-review.md
allowed_tools:
- Read
- Glob
@ -301,22 +346,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: frontend_review (Frontend Expert Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-frontend-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the changes above from the frontend development perspective.
**Review Criteria:**
@ -359,7 +393,7 @@ steps:
```
rules:
- condition: Frontend design is sound with no issues
next: ai_review
next: security_review
- condition: Frontend design issues found
next: fix_frontend
@ -375,21 +409,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_frontend
## Frontend Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the frontend expert.
Areas of concern:
@ -405,122 +427,12 @@ steps:
- condition: Unable to proceed with fixes
next: plan
# ===========================================
# Phase 4: AI Review
# ===========================================
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: ai_review (AI-Generated Code Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/05-ai-review.md
## Original User Request (Initial request from workflow start)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the code for AI-specific issues:
- Assumption validation
- Plausible but wrong patterns
- Context fit with existing codebase
- Scope creep detection
**Report output:** Output to the `Report File` specified above.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# AI-Generated Code Review
## Result: APPROVE / REJECT
## Summary
{One sentence summarizing result}
## Verified Items
| Aspect | Result | Notes |
|--------|--------|-------|
| Assumption validity | ✅ | - |
| API/Library existence | ✅ | - |
| Context fit | ✅ | - |
| Scope | ✅ | - |
## Issues (if REJECT)
| # | Category | Location | Issue |
|---|----------|----------|-------|
| 1 | Hallucinated API | `src/file.ts:23` | Non-existent method |
```
**Cognitive load reduction rules:**
- No issues → Summary 1 line + check table only (10 lines or less)
- Issues found → + Issues in table format (25 lines or less)
rules:
- condition: No AI-specific issues found
next: security_review
- condition: AI-specific issues detected
next: ai_fix
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: ai_fix
## AI Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request from workflow start - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Address the AI Reviewer's feedback.
Focus on:
- Correcting incorrect assumptions
- Fixing plausible-but-wrong implementations
- Aligning with existing codebase patterns
- Removing scope creep
pass_previous_response: true
rules:
- condition: AI Reviewer's issues have been fixed
next: ai_review
- condition: Unable to proceed with fixes
next: plan
# ===========================================
# Phase 5: Security Review
# ===========================================
- name: security_review
agent: ~/.takt/agents/expert/security-reviewer.md
report: 06-security-review.md
allowed_tools:
- Read
- Glob
@ -529,22 +441,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: security_review (Security Expert Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/06-security-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the changes above from the security perspective.
**Review Criteria:**
@ -600,21 +501,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_security
## Security Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the security expert.
Security issues should be addressed with highest priority.
@ -645,6 +534,7 @@ steps:
# ===========================================
- name: qa_review
agent: ~/.takt/agents/expert/qa-reviewer.md
report: 07-qa-review.md
allowed_tools:
- Read
- Glob
@ -653,22 +543,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: qa_review (QA Expert Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/07-qa-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the changes above from the quality assurance perspective.
**Review Criteria:**
@ -724,21 +603,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_qa
## QA Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the QA expert.
Areas of concern:
@ -773,6 +640,9 @@ steps:
# ===========================================
- name: supervise
agent: ~/.takt/agents/expert/supervisor.md
report:
- Validation: 08-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -781,18 +651,6 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: supervise (Final Review)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/08-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
@ -800,13 +658,12 @@ steps:
## Previous Reviews Summary
Reaching this step means all the following reviews have been APPROVED:
- AI Review: APPROVED
- CQRS+ES Review: APPROVED
- Frontend Review: APPROVED
- AI Review: APPROVED
- Security Review: APPROVED
- QA Review: APPROVED
## Instructions
Run tests, verify the build, and perform final approval.
**Workflow Overall Review:**
@ -863,9 +720,9 @@ steps:
## Review Results
| Review | Result |
|--------|--------|
| AI Review | ✅ APPROVE |
| CQRS+ES | ✅ APPROVE |
| Frontend | ✅ APPROVE |
| AI Review | ✅ APPROVE |
| Security | ✅ APPROVE |
| QA | ✅ APPROVE |
| Supervisor | ✅ APPROVE |
@ -894,21 +751,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_supervisor
## Supervisor Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the supervisor.
The supervisor has identified issues from a big-picture perspective.

406
resources/global/en/workflows/expert.yaml
View File

@ -2,13 +2,16 @@
# Review workflow with Architecture, Frontend, Security, and QA experts
#
# Flow:
# plan -> implement -> architect_review -> frontend_review -> ai_review -> security_review -> qa_review -> supervise -> COMPLETE
# ↓ ↓ ↓ ↓ ↓ ↓
# fix_architect fix_frontend ai_fix fix_security fix_qa fix_supervisor
# plan -> implement -> ai_review -> architect_review -> frontend_review -> security_review -> qa_review -> supervise -> COMPLETE
# ↓ ↓ ↓ ↓ ↓ ↓ ↓
# ai_fix fix_architect fix_frontend fix_security fix_qa fix_supervisor
#
# AI review runs immediately after implementation to catch AI-specific issues early,
# before expert reviews begin.
#
# Fix destination is determined by Coder based on change impact:
# - fix_security: MINOR→security_review, MAJOR→architect_review
# - fix_qa: MINOR→qa_review, SECURITY→security_review, MAJOR→architect_review
# - fix_security: MINOR->security_review, MAJOR->architect_review
# - fix_qa: MINOR->qa_review, SECURITY->security_review, MAJOR->architect_review
#
# Template Variables:
# {iteration} - Workflow-wide turn count (total steps executed across all agents)
@ -23,7 +26,7 @@
name: expert
description: Architecture, Frontend, Security, QA Expert Review
max_iterations: 20
max_iterations: 30
initial_step: plan
@ -33,6 +36,7 @@ steps:
# ===========================================
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -42,20 +46,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: plan (Task Analysis)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (when returned from implement)
{previous_response}
## Instructions
Analyze the task and create an implementation plan.
**Note:** If returned from implement step (Previous Response exists),
@ -66,7 +59,7 @@ steps:
2. Identify impact scope
3. Decide implementation approach
**Report output:** Output to the `Report File` specified above.
**Report output:** Output to `.takt/reports/{report_dir}/00-plan.md`.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
@ -103,6 +96,9 @@ steps:
# ===========================================
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -113,26 +109,12 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
Follow the plan from the plan step and implement.
Refer to the plan report (00-plan.md) and proceed with implementation.
**Report output:** Output to the `Report Files` specified above.
**Report output:** Output to the report files specified below.
- Scope: `.takt/reports/{report_dir}/01-coder-scope.md`
- Decisions: `.takt/reports/{report_dir}/02-coder-decisions.md`
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
@ -167,15 +149,105 @@ steps:
```
rules:
- condition: Implementation is complete
next: architect_review
next: ai_review
- condition: Cannot proceed with implementation
next: plan
# ===========================================
# Phase 2: Architecture Review
# Phase 2: AI Review
# ===========================================
- name: architect_review
agent: ~/.takt/agents/default/architecture-reviewer.md
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Git Diff
```diff
{git_diff}
```
Review the code for AI-specific issues:
- Assumption validation
- Plausible but wrong patterns
- Context fit with existing codebase
- Scope creep detection
**Report output:** Output to `.takt/reports/{report_dir}/03-ai-review.md`.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# AI-Generated Code Review
## Result: APPROVE / REJECT
## Summary
{One sentence summarizing result}
## Verified Items
| Aspect | Result | Notes |
|--------|--------|-------|
| Assumption validity | ✅ | - |
| API/Library existence | ✅ | - |
| Context fit | ✅ | - |
| Scope | ✅ | - |
## Issues (if REJECT)
| # | Category | Location | Issue |
|---|----------|----------|-------|
| 1 | Hallucinated API | `src/file.ts:23` | Non-existent method |
```
**Cognitive load reduction rules:**
- No issues -> Summary 1 line + check table only (10 lines or less)
- Issues found -> + Issues in table format (25 lines or less)
rules:
- condition: No AI-specific issues found
next: architect_review
- condition: AI-specific issues detected
next: ai_fix
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
instruction_template: |
## AI Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
**Important**: Address the AI Reviewer's feedback.
Focus on:
- Correcting incorrect assumptions
- Fixing plausible-but-wrong implementations
- Aligning with existing codebase patterns
- Removing scope creep
pass_previous_response: true
rules:
- condition: AI Reviewer's issues have been fixed
next: ai_review
- condition: Unable to proceed with fixes
next: plan
# ===========================================
# Phase 3: Architecture Review
# ===========================================
- name: architect_review
agent: ~/.takt/agents/default/architecture-reviewer.md
report: 04-architect-review.md
allowed_tools:
- Read
- Glob
@ -184,22 +256,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: architect_review (Architecture Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-architect-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Focus on **architecture and design** review.
**Review Criteria:**
@ -210,7 +271,7 @@ steps:
- Dead code
- Call chain verification
**Report output:** Output to the `Report File` specified above.
**Report output:** Output to `.takt/reports/{report_dir}/04-architect-review.md`.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
@ -241,9 +302,9 @@ steps:
```
**Cognitive load reduction rules:**
- APPROVE + no issues → Summary only (5 lines or less)
- APPROVE + minor suggestions → Summary + suggestions (15 lines or less)
- REJECT → Issues in table format (30 lines or less)
- APPROVE + no issues -> Summary only (5 lines or less)
- APPROVE + minor suggestions -> Summary + suggestions (15 lines or less)
- REJECT -> Issues in table format (30 lines or less)
rules:
- condition: No architecture or design issues found
next: frontend_review
@ -265,21 +326,9 @@ steps:
- WebFetch
permission_mode: acceptEdits
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_architect
## Architect Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Address the Architect's feedback.
"Original User Request" is for reference; it's not the latest instruction.
Review the session conversation history and fix the Architect's issues.
@ -291,10 +340,11 @@ steps:
next: plan
# ===========================================
# Phase 3: Frontend Review
# Phase 4: Frontend Review
# ===========================================
- name: frontend_review
agent: ~/.takt/agents/expert/frontend-reviewer.md
report: 05-frontend-review.md
allowed_tools:
- Read
- Glob
@ -303,22 +353,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: frontend_review (Frontend Expert Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-frontend-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the changes above from the frontend development perspective.
**Review Criteria:**
@ -332,7 +371,7 @@ steps:
**Note**: If this project does not include frontend code,
approve and proceed to the next step.
**Report output:** Output to the `Report File` specified above.
**Report output:** Output to `.takt/reports/{report_dir}/05-frontend-review.md`.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
@ -361,7 +400,7 @@ steps:
```
rules:
- condition: Frontend design is sound with no issues
next: ai_review
next: security_review
- condition: Frontend design issues found
next: fix_frontend
@ -377,21 +416,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_frontend
## Frontend Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the frontend expert.
Areas of concern:
@ -407,122 +434,12 @@ steps:
- condition: Unable to proceed with fixes
next: plan
# ===========================================
# Phase 4: AI Review
# ===========================================
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: ai_review (AI-Generated Code Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/05-ai-review.md
## Original User Request (Initial request from workflow start)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the code for AI-specific issues:
- Assumption validation
- Plausible but wrong patterns
- Context fit with existing codebase
- Scope creep detection
**Report output:** Output to the `Report File` specified above.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# AI-Generated Code Review
## Result: APPROVE / REJECT
## Summary
{One sentence summarizing result}
## Verified Items
| Aspect | Result | Notes |
|--------|--------|-------|
| Assumption validity | ✅ | - |
| API/Library existence | ✅ | - |
| Context fit | ✅ | - |
| Scope | ✅ | - |
## Issues (if REJECT)
| # | Category | Location | Issue |
|---|----------|----------|-------|
| 1 | Hallucinated API | `src/file.ts:23` | Non-existent method |
```
**Cognitive load reduction rules:**
- No issues → Summary 1 line + check table only (10 lines or less)
- Issues found → + Issues in table format (25 lines or less)
rules:
- condition: No AI-specific issues found
next: security_review
- condition: AI-specific issues detected
next: ai_fix
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: ai_fix
## AI Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request from workflow start - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Address the AI Reviewer's feedback.
Focus on:
- Correcting incorrect assumptions
- Fixing plausible-but-wrong implementations
- Aligning with existing codebase patterns
- Removing scope creep
pass_previous_response: true
rules:
- condition: AI Reviewer's issues have been fixed
next: ai_review
- condition: Unable to proceed with fixes
next: plan
# ===========================================
# Phase 5: Security Review
# ===========================================
- name: security_review
agent: ~/.takt/agents/expert/security-reviewer.md
report: 06-security-review.md
allowed_tools:
- Read
- Glob
@ -531,22 +448,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: security_review (Security Expert Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/06-security-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the changes above from the security perspective.
**Review Criteria:**
@ -556,7 +462,7 @@ steps:
- Encryption appropriateness
- OWASP Top 10
**Report output:** Output to the `Report File` specified above.
**Report output:** Output to `.takt/reports/{report_dir}/06-security-review.md`.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
@ -602,21 +508,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_security
## Security Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the security expert.
Security issues should be addressed with highest priority.
@ -647,6 +541,7 @@ steps:
# ===========================================
- name: qa_review
agent: ~/.takt/agents/expert/qa-reviewer.md
report: 07-qa-review.md
allowed_tools:
- Read
- Glob
@ -655,22 +550,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: qa_review (QA Expert Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/07-qa-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the changes above from the quality assurance perspective.
**Review Criteria:**
@ -681,7 +565,7 @@ steps:
- Logging and monitoring
- Maintainability
**Report output:** Output to the `Report File` specified above.
**Report output:** Output to `.takt/reports/{report_dir}/07-qa-review.md`.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
@ -726,21 +610,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_qa
## QA Review Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the QA expert.
Areas of concern:
@ -775,6 +647,9 @@ steps:
# ===========================================
- name: supervise
agent: ~/.takt/agents/expert/supervisor.md
report:
- Validation: 08-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -783,18 +658,6 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: supervise (Final Review)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/08-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
@ -808,7 +671,6 @@ steps:
- Security Review: APPROVED
- QA Review: APPROVED
## Instructions
Run tests, verify the build, and perform final approval.
**Workflow Overall Review:**
@ -819,7 +681,9 @@ steps:
**Review Reports:** Read all reports in Report Directory and
check for any unaddressed improvement suggestions.
**Report output:** Output to the `Report Files` specified above.
**Report output:** Output to the report files specified below.
- Validation: `.takt/reports/{report_dir}/08-supervisor-validation.md`
- Summary: `.takt/reports/{report_dir}/summary.md`
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
@ -896,21 +760,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: fix_supervisor
## Supervisor Feedback (This is the latest instruction - prioritize this)
{previous_response}
## Original User Request (Initial request - for reference)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**Important**: Fix the issues pointed out by the supervisor.
The supervisor has identified issues from a big-picture perspective.

214
resources/global/en/workflows/simple.yaml
View File

@ -1,19 +1,24 @@
# Simple TAKT Workflow
# Plan -> Coder -> Architect Review -> AI Review -> Supervisor Approval
# (Simplified version of default - removed improve, fix, ai_fix, security_review, security_fix)
# Plan -> Implement -> AI Review -> Architect Review -> Supervisor Approval
#
# Template Variables:
# {iteration} - Workflow-wide turn count (total steps executed across all agents)
# {max_iterations} - Maximum iterations allowed for the workflow
# {step_iteration} - Per-step iteration count (how many times THIS step has been executed)
# Template Variables (auto-injected by engine):
# {iteration} - Workflow-wide turn count
# {max_iterations} - Maximum iterations allowed
# {step_iteration} - Per-step iteration count
# {task} - Original user request
# {previous_response} - Output from the previous step
# {git_diff} - Current uncommitted changes (git diff)
# {user_inputs} - Accumulated user inputs during workflow
# {report_dir} - Report directory name (e.g., "20250126-143052-task-summary")
# {report_dir} - Report directory name
#
# Auto-injected sections (do NOT include in instruction_template):
# ## Workflow Context - iteration, step_iteration, report info
# ## User Request - {task}
# ## Previous Response - {previous_response}
# ## Additional User Inputs - {user_inputs}
name: simple
description: Simplified development workflow (plan -> implement -> review -> ai_review -> supervise)
description: Simplified development workflow (plan -> implement -> ai_review -> review -> supervise)
max_iterations: 20
@ -22,6 +27,7 @@ initial_step: plan
steps:
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -39,20 +45,9 @@ steps:
next: ABORT
pass_previous_response: true
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: plan (Task Analysis)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (when returned from implement)
{previous_response}
## Instructions
Analyze the task and create an implementation plan.
**Note:** If returned from implement step (Previous Response exists),
@ -91,6 +86,9 @@ steps:
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -103,26 +101,10 @@ steps:
permission_mode: acceptEdits
rules:
- condition: Implementation complete
next: review
next: ai_review
- condition: Cannot proceed, insufficient info
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
Follow the plan from the plan step and implement.
Refer to the plan report (00-plan.md) and proceed with implementation.
@ -160,8 +142,67 @@ steps:
- **Reason**: {Why this option was chosen}
```
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
rules:
- condition: No AI-specific issues
next: review
- condition: AI-specific issues found
next: plan
instruction_template: |
## Git Diff
```diff
{git_diff}
```
Review the code for AI-specific issues:
- Assumption validation
- Plausible but wrong patterns
- Context fit with existing codebase
- Scope creep detection
**Report output:** Output to the `Report File` specified above.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# AI-Generated Code Review
## Result: APPROVE / REJECT
## Summary
{One sentence summarizing result}
## Verified Items
| Aspect | Result | Notes |
|--------|--------|-------|
| Assumption validity | ✅ | - |
| API/Library existence | ✅ | - |
| Context fit | ✅ | - |
| Scope | ✅ | - |
## Issues (if REJECT)
| # | Category | Location | Issue |
|---|----------|----------|-------|
| 1 | Hallucinated API | `src/file.ts:23` | Non-existent method |
```
**Cognitive load reduction rules:**
- No issues -> Summary 1 line + check table only (10 lines or less)
- Issues found -> + Issues in table format (25 lines or less)
- name: review
agent: ~/.takt/agents/default/architecture-reviewer.md
report: 04-architect-review.md
allowed_tools:
- Read
- Glob
@ -171,27 +212,16 @@ steps:
- WebFetch
rules:
- condition: No issues found
next: ai_review
next: supervise
- condition: Structural fix required
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: review (Architecture Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-architect-review.md
## Original User Request (Initial request from workflow start)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Focus on **architecture and design** review. Do NOT review AI-specific issues (that's the next step).
Focus on **architecture and design** review. Do NOT review AI-specific issues (that's already done).
Review the changes and provide feedback.
@ -230,76 +260,11 @@ steps:
- APPROVE + minor suggestions -> Summary + suggestions (15 lines or less)
- REJECT -> Issues in table format (30 lines or less)
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
rules:
- condition: No AI-specific issues
next: supervise
- condition: AI-specific issues found
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: ai_review (AI-Generated Code Review)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-ai-review.md
## Original User Request (Initial request from workflow start)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Review the code for AI-specific issues:
- Assumption validation
- Plausible but wrong patterns
- Context fit with existing codebase
- Scope creep detection
**Report output:** Output to the `Report File` specified above.
- If file does not exist: Create new file
- If file exists: Append with `## Iteration {step_iteration}` section
**Report format:**
```markdown
# AI-Generated Code Review
## Result: APPROVE / REJECT
## Summary
{One sentence summarizing result}
## Verified Items
| Aspect | Result | Notes |
|--------|--------|-------|
| Assumption validity | ✅ | - |
| API/Library existence | ✅ | - |
| Context fit | ✅ | - |
| Scope | ✅ | - |
## Issues (if REJECT)
| # | Category | Location | Issue |
|---|----------|----------|-------|
| 1 | Hallucinated API | `src/file.ts:23` | Non-existent method |
```
**Cognitive load reduction rules:**
- No issues -> Summary 1 line + check table only (10 lines or less)
- Issues found -> + Issues in table format (25 lines or less)
- name: supervise
agent: ~/.takt/agents/default/supervisor.md
report:
- Validation: 05-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -314,24 +279,11 @@ steps:
- condition: Requirements unmet, tests failing
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations} (workflow-wide)
- Step Iteration: {step_iteration} (times this step has run)
- Step: supervise (final verification)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/05-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
Run tests, verify the build, and perform final approval.
**Workflow Overall Review:**
@ -388,8 +340,8 @@ steps:
## Review Results
| Review | Result |
|--------|--------|
| Architect | ✅ APPROVE |
| AI Review | ✅ APPROVE |
| Architect | ✅ APPROVE |
| Supervisor | ✅ APPROVE |
## Verification Commands

49
resources/global/ja/agents/default/ai-antipattern-reviewer.md
View File

@ -149,7 +149,54 @@ AIは自信を持って間違える——もっともらしく見えるが動か
2. 各フォールバックに正当な理由があるか確認
3. 理由なしのフォールバックが1つでもあれば REJECT
### 8. 決定トレーサビリティレビュー
### 8. 未使用コードの検出
**AIは「将来の拡張性」「対称性」「念のため」で不要なコードを生成しがちである。現時点で呼ばれていないコードは削除する。**
| 判定 | 基準 |
|------|------|
| **REJECT** | 現在どこからも呼ばれていないpublic関数・メソッド |
| **REJECT** | 「対称性のため」に作られたが使われていないsetter/getter |
| **REJECT** | 将来の拡張のために用意されたインターフェースやオプション |
| **REJECT** | exportされているが、grep で使用箇所が見つからない |
| OK | フレームワークが暗黙的に呼び出す（ライフサイクルフック等） |
| OK | 公開パッケージのAPIとして意図的に公開している |
**検証アプローチ:**
1. 変更・削除されたコードを参照している箇所がないか grep で確認
2. 公開モジュール（index ファイル等）のエクスポート一覧と実体が一致しているか確認
3. 新規追加されたコードに対応する古いコードが残っていないか確認
### 9. 不要な後方互換コードの検出
**AIは「後方互換のために」不要なコードを残しがちである。これを見逃さない。**
削除すべき後方互換コード:
| パターン | 例 | 判定 |
|---------|-----|------|
| deprecated + 使用箇所なし | `@deprecated` アノテーション付きで誰も使っていない | **即削除** |
| 新APIと旧API両方存在 | 新関数があるのに旧関数も残っている | 旧を**削除** |
| 移行済みのラッパー | 互換のために作ったが移行完了済み | **削除** |
| コメントで「将来削除」 | `// TODO: remove after migration` が放置 | **今すぐ削除** |
| Proxy/アダプタの過剰使用 | 後方互換のためだけに複雑化 | **シンプルに置換** |
残すべき後方互換コード:
| パターン | 例 | 判定 |
|---------|-----|------|
| 外部公開API | npm パッケージのエクスポート | 慎重に検討 |
| 設定ファイル互換 | 旧形式の設定を読める | メジャーバージョンまで維持 |
| データ移行中 | DBスキーマ移行の途中 | 移行完了まで維持 |
**判断基準:**
1. **使用箇所があるか？** → grep/検索で確認。なければ削除
2. **外部に公開しているか？** → 内部のみなら即削除可能
3. **移行は完了したか？** → 完了なら削除
**AIが「後方互換のため」と言ったら疑う。** 本当に必要か確認せよ。
### 10. 決定トレーサビリティレビュー
**Coderの決定ログが妥当か検証する。**

96
resources/global/ja/agents/default/architecture-reviewer.md
View File

@ -191,73 +191,6 @@ if (status === 'interrupted') {
}
```
**フォールバック値の乱用の判定基準:**
フォールバック値（`??`, `||`, デフォルト引数）は「値が無いケース」を握りつぶす。本来エラーにすべき箇所を隠してしまう。
| 判定 | 基準 |
|------|------|
| **REJECT** | 値が無い状態がバグであるのにフォールバックで隠している |
| **REJECT** | `'unknown'`, `'default'`, `''`, `0` など意味のない値でごまかしている |
| **REJECT** | 全呼び出し元がフォールバックに頼り、本来の値を渡していない |
| OK | 外部入力（ユーザー入力、API応答）に対する防御的デフォルト |
| OK | オプショナルな設定項目に対する合理的な初期値 |
```typescript
// ❌ REJECT - バグを隠すフォールバック
const userName = user.name ?? 'unknown'; // name が無いのはデータ不整合
const stepName = step?.name ?? 'default'; // step が無いのは呼び出し元のバグ
// ❌ REJECT - 全呼び出し元が省略しているオプション
function runStep(step: Step, options?: { maxRetries?: number }) {
const retries = options?.maxRetries ?? 3; // 全呼び出し元が options を渡していない
}
// ✅ OK - ユーザー設定のオプショナルなデフォルト
const logLevel = config.logLevel ?? 'info'; // 設定ファイルに無ければ info
const language = userPreference.lang ?? 'en'; // 未設定なら英語
// ✅ OK - 外部APIの防御的デフォルト
const displayName = apiResponse.nickname ?? apiResponse.email; // ニックネーム未設定の場合
```
**未使用コードの判定基準:**
AIは「将来の拡張性」「対称性」「念のため」で不要なコードを生成しがちである。現時点で呼ばれていないコードは削除する。
| 判定 | 基準 |
|------|------|
| **REJECT** | 現在どこからも呼ばれていないpublic関数・メソッド |
| **REJECT** | 「対称性のため」に作られたが使われていないsetter/getter |
| **REJECT** | 将来の拡張のために用意されたインターフェースやオプション |
| **REJECT** | exportされているが、grep で使用箇所が見つからない |
| OK | フレームワークが暗黙的に呼び出す（ライフサイクルフック等） |
| OK | 公開パッケージのAPIとして意図的に公開している |
```typescript
// ❌ REJECT - 「対称性のため」のsetter（getしか使っていない）
class WorkflowState {
private _status: Status;
getStatus(): Status { return this._status; }
setStatus(s: Status) { this._status = s; } // 誰も呼んでいない
}
// ❌ REJECT - 「将来の拡張」のためのオプション
interface EngineOptions {
maxIterations: number;
enableParallel?: boolean; // 未実装。どこからも参照されていない
pluginHooks?: PluginHook[]; // 未実装。プラグイン機構は存在しない
}
// ❌ REJECT - exportされているが使われていない
export function formatStepName(name: string): string { ... } // grep 結果: 0件
// ✅ OK - フレームワークが呼ぶ
class MyComponent extends React.Component {
componentDidMount() { ... } // Reactが呼ぶ
}
```
**状態の直接変更の判定基準:**
オブジェクトや配列を直接変更すると、変更の追跡が困難になり、予期しない副作用を生む。常にスプレッド演算子やイミュータブルな操作で新しいオブジェクトを返す。
@ -379,35 +312,6 @@ function createUser(data: UserData) {
}
```
### 7. 不要な後方互換コードの検出
**AIは「後方互換のために」不要なコードを残しがちである。これを見逃さない。**
削除すべき後方互換コード:
| パターン | 例 | 判定 |
|---------|-----|------|
| deprecated + 使用箇所なし | `@deprecated` アノテーション付きで誰も使っていない | **即削除** |
| 新APIと旧API両方存在 | 新関数があるのに旧関数も残っている | 旧を**削除** |
| 移行済みのラッパー | 互換のために作ったが移行完了済み | **削除** |
| コメントで「将来削除」 | `// TODO: remove after migration` が放置 | **今すぐ削除** |
| Proxy/アダプタの過剰使用 | 後方互換のためだけに複雑化 | **シンプルに置換** |
残すべき後方互換コード:
| パターン | 例 | 判定 |
|---------|-----|------|
| 外部公開API | npm パッケージのエクスポート | 慎重に検討 |
| 設定ファイル互換 | 旧形式の設定を読める | メジャーバージョンまで維持 |
| データ移行中 | DBスキーマ移行の途中 | 移行完了まで維持 |
**判断基準:**
1. **使用箇所があるか？** → grep/検索で確認。なければ削除
2. **外部に公開しているか？** → 内部のみなら即削除可能
3. **移行は完了したか？** → 完了なら削除
**AIが「後方互換のため」と言ったら疑う。** 本当に必要か確認せよ。
### 7. その場しのぎの検出
**「とりあえず動かす」ための妥協を見逃さない。**

334
resources/global/ja/workflows/default.yaml
View File

@ -1,7 +1,7 @@
# Default TAKT Workflow
# Plan -> Coder -> Architect Review -> AI Review -> Security Review -> Supervisor Approval
# Plan -> Implement -> AI Review -> Architect Review -> Security Review -> Supervisor Approval
#
# Template Variables:
# Template Variables (auto-injected by buildInstruction):
# {iteration} - Workflow-wide turn count (total steps executed across all agents)
# {max_iterations} - Maximum iterations allowed for the workflow
# {step_iteration} - Per-step iteration count (how many times THIS step has been executed)
@ -14,13 +14,14 @@
name: default
description: Standard development workflow with planning and specialized reviews
max_iterations: 20
max_iterations: 30
initial_step: plan
steps:
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -44,20 +45,9 @@ steps:
- {質問2}
pass_previous_response: true
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: plan (タスク分析)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (implementからの差し戻し時)
{previous_response}
## Instructions
タスクを分析し、実装方針を立ててください。
**注意:** Previous Responseがある場合は差し戻しのため、
@ -68,7 +58,7 @@ steps:
2. 影響範囲を特定する
3. 実装アプローチを決める
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** Report File に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -96,6 +86,9 @@ steps:
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -108,26 +101,10 @@ steps:
permission_mode: acceptEdits
rules:
- condition: 実装完了
next: review
next: ai_review
- condition: 判断できない、情報不足
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
planステップで立てた計画に従って実装してください。
計画レポート（00-plan.md）を参照し、実装を進めてください。
@ -136,7 +113,7 @@ steps:
- 既存コードを変更した場合は該当するテストを更新
- テストファイルの配置: プロジェクトの規約に従う（例: `__tests__/`, `*.test.ts`）
**レポート出力:** 上記の `Report Files` に出力してください。
**レポート出力:** Report Files に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -170,8 +147,96 @@ steps:
- **理由**: {選んだ理由}
```
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
rules:
- condition: AI特有の問題なし
next: review
- condition: AI特有の問題あり
next: ai_fix
instruction_template: |
## Git Diff
```diff
{git_diff}
```
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レポート出力:** Report File に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# AI生成コードレビュー
## 結果: APPROVE / REJECT
## サマリー
{1文で結果を要約}
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
permission_mode: acceptEdits
rules:
- condition: AI問題の修正完了
next: review
- condition: 判断できない、情報不足
next: plan
instruction_template: |
## AI Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
**重要**: AI Reviewerのフィードバックに対応してください。
以下に集中してください:
- 間違った仮定の修正
- もっともらしいが間違っている実装の修正
- 既存コードベースのパターンとの整合
- スコープクリープの除去
pass_previous_response: true
- name: review
agent: ~/.takt/agents/default/architecture-reviewer.md
report: 04-architect-review.md
allowed_tools:
- Read
- Glob
@ -181,28 +246,17 @@ steps:
- WebFetch
rules:
- condition: 問題なし
next: ai_review
next: security_review
- condition: 改善すべき点がある（軽微）
next: improve
- condition: 構造的な修正が必要
next: fix
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: review (アーキテクチャレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-architect-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
**アーキテクチャと設計**のレビューに集中してください。AI特有の問題はレビューしないでください（次のステップで行います）。
変更をレビューしてフィードバックを提供してください。
@ -215,7 +269,7 @@ steps:
- デッドコード
- 呼び出しチェーン検証
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** Report File に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -268,21 +322,9 @@ steps:
- condition: 判断できない、情報不足
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: improve
## Architect Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: Architectの改善提案に対応してください。
これらは軽微な改善であり、設計上の大きな問題ではありません。
@ -293,75 +335,7 @@ steps:
- コードの整理
pass_previous_response: true
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
rules:
- condition: AI特有の問題なし
next: security_review
- condition: AI特有の問題あり
next: ai_fix
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: ai_review (AI生成コードレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-ai-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レポート出力:** 上記の `Report File` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# AI生成コードレビュー
## 結果: APPROVE / REJECT
## サマリー
{1文で結果を要約}
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
- name: ai_fix
- name: fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
@ -374,36 +348,21 @@ steps:
- WebFetch
permission_mode: acceptEdits
rules:
- condition: AI問題の修正完了
- condition: Architectの指摘を修正完了
next: review
- condition: 判断できない、情報不足
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: ai_fix
## AI Review Feedback (これが最新の指示です - 優先して対応してください)
## Architect Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: AI Reviewerのフィードバックに対応してください。
以下に集中してください:
- 間違った仮定の修正
- もっともらしいが間違っている実装の修正
- 既存コードベースのパターンとの整合
- スコープクリープの除去
**重要**: Architectのフィードバックに対応してください。
セッションの会話履歴を確認し、Architectの指摘事項を修正してください。
pass_previous_response: true
- name: security_review
agent: ~/.takt/agents/default/security-reviewer.md
report: 05-security-review.md
allowed_tools:
- Read
- Glob
@ -417,29 +376,18 @@ steps:
- condition: 脆弱性があり修正が必要
next: security_fix
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: security_review
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/05-security-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
変更に対してセキュリティレビューを行ってください。以下の脆弱性を確認してください:
- インジェクション攻撃（SQL, コマンド, XSS）
- 認証・認可の問題
- データ露出リスク
- 暗号化の弱点
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** Report File に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -491,65 +439,18 @@ steps:
- condition: 判断できない、情報不足
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: security_fix
## Security Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: セキュリティレビューで指摘された脆弱性を修正してください。
セキュリティの問題は最優先で対応が必要です。
pass_previous_response: true
- name: fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
permission_mode: acceptEdits
rules:
- condition: Architectの指摘を修正完了
next: review
- condition: 判断できない、情報不足
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix
## Architect Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: Architectのフィードバックに対応してください。
「Original User Request」は参考情報であり、最新の指示ではありません。
セッションの会話履歴を確認し、Architectの指摘事項を修正してください。
pass_previous_response: true
- name: supervise
agent: ~/.takt/agents/default/supervisor.md
report:
- Validation: 06-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -564,24 +465,11 @@ steps:
- condition: 要求未達成、テスト失敗、ビルドエラー
next: plan
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: supervise (final verification)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/06-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
テスト実行、ビルド確認、最終承認を行ってください。
**ワークフロー全体の確認:**
@ -592,7 +480,7 @@ steps:
**レポートの確認:** Report Directory内の全レポートを読み、
未対応の改善提案がないか確認してください。
**レポート出力:** 上記の `Report Files` に出力してください。
**レポート出力:** Report Files に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記

412
resources/global/ja/workflows/expert-cqrs.yaml
View File

@ -2,9 +2,9 @@
# CQRS+ES、フロントエンド、セキュリティ、QAの専門家によるレビューワークフロー
#
# フロー:
# plan -> implement -> architect_review -> cqrs_es_review -> frontend_review -> ai_review -> security_review -> qa_review -> supervise -> COMPLETE
# ↓ ↓ ↓ ↓ ↓ ↓ ↓
# fix_architect fix_cqrs_es fix_frontend ai_fix fix_security fix_qa fix_supervisor
# plan -> implement -> ai_review -> cqrs_es_review -> frontend_review -> security_review -> qa_review -> supervise -> COMPLETE
# ↓ ↓ ↓ ↓ ↓ ↓
# ai_fix fix_cqrs_es fix_frontend fix_security fix_qa fix_supervisor
#
# 修正時の戻り先はCoderが判断:
# - fix_security: MINOR→security_review, MAJOR→cqrs_es_review
@ -23,7 +23,7 @@
name: expert-cqrs
description: CQRS+ES・フロントエンド・セキュリティ・QA専門家レビュー
max_iterations: 20
max_iterations: 30
initial_step: plan
@ -33,6 +33,7 @@ steps:
# ===========================================
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -42,20 +43,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: plan (タスク分析)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (implementからの差し戻し時)
{previous_response}
## Instructions
タスクを分析し、実装方針を立ててください。
**注意:** Previous Responseがある場合は差し戻しのため、
@ -66,7 +56,7 @@ steps:
2. 影響範囲を特定する
3. 実装アプローチを決める
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** .takt/reports/{report_dir}/00-plan.md に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -103,6 +93,9 @@ steps:
# ===========================================
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -113,26 +106,12 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
planステップで立てた計画に従って実装してください。
計画レポート（00-plan.md）を参照し、実装を進めてください。
**レポート出力:** 上記の `Report Files` に出力してください。
**レポート出力:** .takt/reports/{report_dir}/ に出力してください。
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -167,15 +146,16 @@ steps:
```
rules:
- condition: 実装が完了した
next: architect_review
next: ai_review
- condition: 実装を進行できない
next: plan
# ===========================================
# Phase 2: Architecture Review
# Phase 2: AI Review
# ===========================================
- name: architect_review
agent: ~/.takt/agents/default/architecture-reviewer.md
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
@ -184,75 +164,54 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: architect_review (アーキテクチャレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-architect-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
**アーキテクチャと設計**のレビューに集中してください。
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レビュー観点:**
- 構造・設計の妥当性
- コード品質
- 変更スコープの適切性
- テストカバレッジ
- デッドコード
- 呼び出しチェーン検証
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** .takt/reports/{report_dir}/03-ai-review.md に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# アーキテクチャレビュー
# AI生成コードレビュー
## 結果: APPROVE / IMPROVE / REJECT
## 結果: APPROVE / REJECT
## サマリー
{1-2文で結果を要約}
{1文で結果を要約}
## 確認した観点
- [x] 構造・設計
- [x] コード品質
- [x] 変更スコープ
- [x] テストカバレッジ
- [x] デッドコード
- [x] 呼び出しチェーン検証
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | 場所 | 問題 | 修正案 |
|---|------|------|--------|
| 1 | `src/file.ts:42` | 問題の説明 | 修正方法 |
## 改善提案（任意・ブロッキングではない）
- {将来的な改善提案}
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- APPROVE + 問題なし → サマリーのみ（5行以内）
- APPROVE + 軽微な提案 → サマリー + 改善提案（15行以内）
- REJECT → 問題点を表形式で（30行以内）
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
rules:
- condition: アーキテクチャと設計に問題がない
- condition: AI特有の問題が見つからない
next: cqrs_es_review
- condition: 軽微な改善が必要だが構造的な問題はない
next: fix_architect
- condition: 構造的な問題があり修正が必要
next: fix_architect
- condition: AI特有の問題が検出された
next: ai_fix
- name: fix_architect
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
@ -263,30 +222,20 @@ steps:
- Bash
- WebSearch
- WebFetch
permission_mode: acceptEdits
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_architect
## Architect Feedback (これが最新の指示です - 優先して対応してください)
## AI Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: Architectのフィードバックに対応してください。
「Original User Request」は参考情報であり、最新の指示ではありません。
セッションの会話履歴を確認し、Architectの指摘事項を修正してください。
**重要**: AI Reviewerのフィードバックに対応してください。
以下に集中してください:
- 間違った仮定の修正
- もっともらしいが間違っている実装の修正
- 既存コードベースのパターンとの整合
- スコープクリープの除去
pass_previous_response: true
rules:
- condition: Architectの指摘に対する修正が完了した
next: architect_review
- condition: AI Reviewerの指摘に対する修正が完了した
next: ai_review
- condition: 修正を進行できない
next: plan
@ -295,6 +244,7 @@ steps:
# ===========================================
- name: cqrs_es_review
agent: ~/.takt/agents/expert-cqrs/cqrs-es-reviewer.md
report: 04-cqrs-es-review.md
allowed_tools:
- Read
- Glob
@ -303,22 +253,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: cqrs_es_review (CQRS+ES専門レビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-cqrs-es-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
CQRS（コマンドクエリ責務分離）とEvent Sourcing（イベントソーシング）の観点から
上記の変更をレビューしてください。
@ -332,7 +271,7 @@ steps:
**注意**: このプロジェクトがCQRS+ESパターンを使用していない場合は、
一般的なドメイン設計の観点からレビューしてください。
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** .takt/reports/{report_dir}/04-cqrs-es-review.md に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -377,21 +316,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_cqrs_es
## CQRS+ES Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: CQRS+ES専門家からの指摘を修正してください。
指摘されたポイント:
@ -408,10 +335,11 @@ steps:
next: plan
# ===========================================
# Phase 3: Frontend Review
# Phase 4: Frontend Review
# ===========================================
- name: frontend_review
agent: ~/.takt/agents/expert/frontend-reviewer.md
report: 05-frontend-review.md
allowed_tools:
- Read
- Glob
@ -420,22 +348,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: frontend_review (フロントエンド専門レビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/05-frontend-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
フロントエンド開発の観点から上記の変更をレビューしてください。
**レビュー観点:**
@ -449,7 +366,7 @@ steps:
**注意**: このプロジェクトがフロントエンドを含まない場合は、
問題なしとして次に進んでください。
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** .takt/reports/{report_dir}/05-frontend-review.md に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -478,7 +395,7 @@ steps:
```
rules:
- condition: フロントエンド設計に問題がない
next: ai_review
next: security_review
- condition: フロントエンド設計に問題がある
next: fix_frontend
@ -494,21 +411,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_frontend
## Frontend Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: フロントエンド専門家からの指摘を修正してください。
指摘されたポイント:
@ -524,122 +429,12 @@ steps:
- condition: 修正を進行できない
next: plan
# ===========================================
# Phase 4: AI Review
# ===========================================
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: ai_review (AI生成コードレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/06-ai-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レポート出力:** 上記の `Report File` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# AI生成コードレビュー
## 結果: APPROVE / REJECT
## サマリー
{1文で結果を要約}
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
rules:
- condition: AI特有の問題が見つからない
next: security_review
- condition: AI特有の問題が検出された
next: ai_fix
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: ai_fix
## AI Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: AI Reviewerのフィードバックに対応してください。
以下に集中してください:
- 間違った仮定の修正
- もっともらしいが間違っている実装の修正
- 既存コードベースのパターンとの整合
- スコープクリープの除去
pass_previous_response: true
rules:
- condition: AI Reviewerの指摘に対する修正が完了した
next: ai_review
- condition: 修正を進行できない
next: plan
# ===========================================
# Phase 5: Security Review
# ===========================================
- name: security_review
agent: ~/.takt/agents/expert/security-reviewer.md
report: 06-security-review.md
allowed_tools:
- Read
- Glob
@ -648,22 +443,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: security_review (セキュリティ専門レビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/07-security-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
セキュリティの観点から上記の変更をレビューしてください。
**レビュー観点:**
@ -673,7 +457,7 @@ steps:
- 暗号化の適切性
- OWASP Top 10
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** .takt/reports/{report_dir}/06-security-review.md に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -719,21 +503,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_security
## Security Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: セキュリティ専門家からの指摘を修正してください。
セキュリティ問題は最優先で対応してください。
@ -764,6 +536,7 @@ steps:
# ===========================================
- name: qa_review
agent: ~/.takt/agents/expert/qa-reviewer.md
report: 07-qa-review.md
allowed_tools:
- Read
- Glob
@ -772,22 +545,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: qa_review (QA専門レビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/08-qa-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
品質保証の観点から上記の変更をレビューしてください。
**レビュー観点:**
@ -798,7 +560,7 @@ steps:
- ログとモニタリング
- 保守性
**レポート出力:** 上記の `Report File` に出力してください。
**レポート出力:** .takt/reports/{report_dir}/07-qa-review.md に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -843,21 +605,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_qa
## QA Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: QA専門家からの指摘を修正してください。
指摘されたポイント:
@ -892,6 +642,9 @@ steps:
# ===========================================
- name: supervise
agent: ~/.takt/agents/expert/supervisor.md
report:
- Validation: 08-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -900,18 +653,6 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: supervise (最終確認)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/09-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
@ -919,14 +660,12 @@ steps:
## Previous Reviews Summary
このステップに到達したということは、以下のレビューがすべてAPPROVEされています：
- Architecture Review: APPROVED
- AI Review: APPROVED
- CQRS+ES Review: APPROVED
- Frontend Review: APPROVED
- AI Review: APPROVED
- Security Review: APPROVED
- QA Review: APPROVED
## Instructions
テスト実行、ビルド確認、最終承認を行ってください。
**ワークフロー全体の確認:**
@ -934,10 +673,12 @@ steps:
2. 各レビューステップの指摘が対応されているか
3. 元のタスク目的が達成されているか
**レポートの確認:** Report Directory内の全レポートを読み、
**レポートの確認:** .takt/reports/{report_dir}/ 内の全レポートを読み、
未対応の改善提案がないか確認してください。
**レポート出力:** 上記の `Report Files` に出力してください。
**レポート出力:**
- Validation: .takt/reports/{report_dir}/08-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
@ -983,10 +724,9 @@ steps:
## レビュー結果
| レビュー | 結果 |
|---------|------|
| Architecture | ✅ APPROVE |
| AI Review | ✅ APPROVE |
| CQRS+ES | ✅ APPROVE |
| Frontend | ✅ APPROVE |
| AI Review | ✅ APPROVE |
| Security | ✅ APPROVE |
| QA | ✅ APPROVE |
| Supervisor | ✅ APPROVE |
@ -1015,21 +755,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_supervisor
## Supervisor Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: 監督者からの指摘を修正してください。
監督者は全体を俯瞰した視点から問題を指摘しています。

377
resources/global/ja/workflows/expert.yaml
View File

@ -2,9 +2,9 @@
# アーキテクチャ、フロントエンド、セキュリティ、QAの専門家によるレビューワークフロー
#
# フロー:
# plan -> implement -> architect_review -> frontend_review -> ai_review -> security_review -> qa_review -> supervise -> COMPLETE
# plan -> implement -> ai_review -> architect_review -> frontend_review -> security_review -> qa_review -> supervise -> COMPLETE
# ↓ ↓ ↓ ↓ ↓ ↓
# fix_architect fix_frontend ai_fix fix_security fix_qa fix_supervisor
# ai_fix fix_architect fix_frontend fix_security fix_qa fix_supervisor
#
# 修正時の戻り先はCoderが判断:
# - fix_security: MINOR→security_review, MAJOR→architect_review
@ -23,7 +23,7 @@
name: expert
description: アーキテクチャ・フロントエンド・セキュリティ・QA専門家レビュー
max_iterations: 20
max_iterations: 30
initial_step: plan
@ -33,6 +33,7 @@ steps:
# ===========================================
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -42,20 +43,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: plan (タスク分析)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (implementからの差し戻し時)
{previous_response}
## Instructions
タスクを分析し、実装方針を立ててください。
**注意:** Previous Responseがある場合は差し戻しのため、
@ -103,6 +93,9 @@ steps:
# ===========================================
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -113,22 +106,6 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
planステップで立てた計画に従って実装してください。
計画レポート（00-plan.md）を参照し、実装を進めてください。
@ -167,15 +144,105 @@ steps:
```
rules:
- condition: 実装が完了した
next: architect_review
next: ai_review
- condition: 実装を進行できない
next: plan
# ===========================================
# Phase 2: Architecture Review
# Phase 2: AI Review
# ===========================================
- name: architect_review
agent: ~/.takt/agents/default/architecture-reviewer.md
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Git Diff
```diff
{git_diff}
```
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レポート出力:** 上記の `Report File` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# AI生成コードレビュー
## 結果: APPROVE / REJECT
## サマリー
{1文で結果を要約}
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
rules:
- condition: AI特有の問題が見つからない
next: architect_review
- condition: AI特有の問題が検出された
next: ai_fix
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
instruction_template: |
## AI Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
**重要**: AI Reviewerのフィードバックに対応してください。
以下に集中してください:
- 間違った仮定の修正
- もっともらしいが間違っている実装の修正
- 既存コードベースのパターンとの整合
- スコープクリープの除去
pass_previous_response: true
rules:
- condition: AI Reviewerの指摘に対する修正が完了した
next: ai_review
- condition: 修正を進行できない
next: plan
# ===========================================
# Phase 3: Architecture Review
# ===========================================
- name: architect_review
agent: ~/.takt/agents/default/architecture-reviewer.md
report: 04-architect-review.md
allowed_tools:
- Read
- Glob
@ -184,22 +251,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: architect_review (アーキテクチャレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-architect-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
**アーキテクチャと設計**のレビューに集中してください。
**レビュー観点:**
@ -265,21 +321,9 @@ steps:
- WebFetch
permission_mode: acceptEdits
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_architect
## Architect Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: Architectのフィードバックに対応してください。
「Original User Request」は参考情報であり、最新の指示ではありません。
セッションの会話履歴を確認し、Architectの指摘事項を修正してください。
@ -291,10 +335,11 @@ steps:
next: plan
# ===========================================
# Phase 3: Frontend Review
# Phase 4: Frontend Review
# ===========================================
- name: frontend_review
agent: ~/.takt/agents/expert/frontend-reviewer.md
report: 05-frontend-review.md
allowed_tools:
- Read
- Glob
@ -303,22 +348,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: frontend_review (フロントエンド専門レビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-frontend-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
フロントエンド開発の観点から上記の変更をレビューしてください。
**レビュー観点:**
@ -361,7 +395,7 @@ steps:
```
rules:
- condition: フロントエンド設計に問題がない
next: ai_review
next: security_review
- condition: フロントエンド設計に問題がある
next: fix_frontend
@ -377,21 +411,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_frontend
## Frontend Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: フロントエンド専門家からの指摘を修正してください。
指摘されたポイント:
@ -407,122 +429,12 @@ steps:
- condition: 修正を進行できない
next: plan
# ===========================================
# Phase 4: AI Review
# ===========================================
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: ai_review (AI生成コードレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/05-ai-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レポート出力:** 上記の `Report File` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# AI生成コードレビュー
## 結果: APPROVE / REJECT
## サマリー
{1文で結果を要約}
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
rules:
- condition: AI特有の問題が見つからない
next: security_review
- condition: AI特有の問題が検出された
next: ai_fix
- name: ai_fix
agent: ~/.takt/agents/default/coder.md
allowed_tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: ai_fix
## AI Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: AI Reviewerのフィードバックに対応してください。
以下に集中してください:
- 間違った仮定の修正
- もっともらしいが間違っている実装の修正
- 既存コードベースのパターンとの整合
- スコープクリープの除去
pass_previous_response: true
rules:
- condition: AI Reviewerの指摘に対する修正が完了した
next: ai_review
- condition: 修正を進行できない
next: plan
# ===========================================
# Phase 5: Security Review
# ===========================================
- name: security_review
agent: ~/.takt/agents/expert/security-reviewer.md
report: 06-security-review.md
allowed_tools:
- Read
- Glob
@ -531,22 +443,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: security_review (セキュリティ専門レビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/06-security-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
セキュリティの観点から上記の変更をレビューしてください。
**レビュー観点:**
@ -602,21 +503,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_security
## Security Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: セキュリティ専門家からの指摘を修正してください。
セキュリティ問題は最優先で対応してください。
@ -647,6 +536,7 @@ steps:
# ===========================================
- name: qa_review
agent: ~/.takt/agents/expert/qa-reviewer.md
report: 07-qa-review.md
allowed_tools:
- Read
- Glob
@ -655,22 +545,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: qa_review (QA専門レビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/07-qa-review.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
品質保証の観点から上記の変更をレビューしてください。
**レビュー観点:**
@ -726,21 +605,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_qa
## QA Review Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: QA専門家からの指摘を修正してください。
指摘されたポイント:
@ -775,6 +642,9 @@ steps:
# ===========================================
- name: supervise
agent: ~/.takt/agents/expert/supervisor.md
report:
- Validation: 08-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -783,18 +653,6 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: supervise (最終確認)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/08-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
@ -802,13 +660,12 @@ steps:
## Previous Reviews Summary
このステップに到達したということは、以下のレビューがすべてAPPROVEされています：
- AI Review: APPROVED
- Architecture Review: APPROVED
- Frontend Review: APPROVED
- AI Review: APPROVED
- Security Review: APPROVED
- QA Review: APPROVED
## Instructions
テスト実行、ビルド確認、最終承認を行ってください。
**ワークフロー全体の確認:**
@ -865,9 +722,9 @@ steps:
## レビュー結果
| レビュー | 結果 |
|---------|------|
| AI Review | ✅ APPROVE |
| Architecture | ✅ APPROVE |
| Frontend | ✅ APPROVE |
| AI Review | ✅ APPROVE |
| Security | ✅ APPROVE |
| QA | ✅ APPROVE |
| Supervisor | ✅ APPROVE |
@ -896,21 +753,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: fix_supervisor
## Supervisor Feedback (これが最新の指示です - 優先して対応してください)
{previous_response}
## Original User Request (ワークフロー開始時の元の要求 - 参考情報)
{task}
## Additional User Inputs
{user_inputs}
## Instructions
**重要**: 監督者からの指摘を修正してください。
監督者は全体を俯瞰した視点から問題を指摘しています。

221
resources/global/ja/workflows/simple.yaml
View File

@ -1,19 +1,19 @@
# Simple TAKT Workflow
# Plan -> Coder -> Architect Review -> AI Review -> Supervisor Approval
# Plan -> Implement -> AI Review -> Architect Review -> Supervisor Approval
# (defaultの簡略版 - improve, fix, ai_fix, security_review, security_fix を削除)
#
# Template Variables:
# Template Variables (auto-injected):
# {iteration} - Workflow-wide turn count (total steps executed across all agents)
# {max_iterations} - Maximum iterations allowed for the workflow
# {step_iteration} - Per-step iteration count (how many times THIS step has been executed)
# {task} - Original user request
# {previous_response} - Output from the previous step
# {task} - Original user request (auto-injected)
# {previous_response} - Output from the previous step (auto-injected)
# {git_diff} - Current uncommitted changes (git diff)
# {user_inputs} - Accumulated user inputs during workflow
# {user_inputs} - Accumulated user inputs during workflow (auto-injected)
# {report_dir} - Report directory name (e.g., "20250126-143052-task-summary")
name: simple
description: Simplified development workflow (plan -> implement -> review -> ai_review -> supervise)
description: Simplified development workflow (plan -> implement -> ai_review -> review -> supervise)
max_iterations: 20
@ -22,6 +22,7 @@ initial_step: plan
steps:
- name: plan
agent: ~/.takt/agents/default/planner.md
report: 00-plan.md
allowed_tools:
- Read
- Glob
@ -31,20 +32,9 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: plan (タスク分析)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/00-plan.md
## User Request
{task}
## Previous Response (implementからの差し戻し時)
{previous_response}
## Instructions
タスクを分析し、実装方針を立ててください。
**注意:** Previous Responseがある場合は差し戻しのため、
@ -55,10 +45,6 @@ steps:
2. 影響範囲を特定する
3. 実装アプローチを決める
**レポート出力:** 上記の `Report File` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# タスク計画
@ -95,6 +81,9 @@ steps:
- name: implement
agent: ~/.takt/agents/default/coder.md
report:
- Scope: 01-coder-scope.md
- Decisions: 02-coder-decisions.md
allowed_tools:
- Read
- Glob
@ -106,29 +95,9 @@ steps:
- WebFetch
permission_mode: acceptEdits
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: implement
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Scope: .takt/reports/{report_dir}/01-coder-scope.md
- Decisions: .takt/reports/{report_dir}/02-coder-decisions.md
## User Request
{task}
## Additional User Inputs
{user_inputs}
## Instructions
planステップで立てた計画に従って実装してください。
計画レポート（00-plan.md）を参照し、実装を進めてください。
**レポート出力:** 上記の `Report Files` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**Scopeレポートフォーマット（実装開始時に作成）:**
```markdown
# 変更スコープ宣言
@ -160,12 +129,13 @@ steps:
```
rules:
- condition: "実装完了"
next: review
next: ai_review
- condition: "判断できない、情報不足"
next: plan
- name: review
agent: ~/.takt/agents/default/architecture-reviewer.md
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
report: 03-ai-review.md
allowed_tools:
- Read
- Glob
@ -174,33 +144,72 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: review (アーキテクチャレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/03-architect-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
**アーキテクチャと設計**のレビューに集中してください。AI特有の問題はレビューしないでください（次のステップで行います）。
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レポートフォーマット:**
```markdown
# AI生成コードレビュー
## 結果: APPROVE / REJECT
## サマリー
{1文で結果を要約}
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
rules:
- condition: "AI特有の問題なし"
next: review
- condition: "AI特有の問題あり"
next: plan
- name: review
agent: ~/.takt/agents/default/architecture-reviewer.md
report: 04-architect-review.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Git Diff
```diff
{git_diff}
```
**アーキテクチャと設計**のレビューに集中してください。AI特有の問題はレビューしないでください（前のステップで完了済み）。
変更をレビューしてフィードバックを提供してください。
**注意:** simpleワークフローではIMPROVE判定は使用しません。
軽微な改善提案がある場合は APPROVE + コメントとしてください。
**レポート出力:** 上記の `Report File` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# アーキテクチャレビュー
@ -230,80 +239,15 @@ steps:
- REJECT → 問題点を表形式で（30行以内）
rules:
- condition: "問題なし"
next: ai_review
- condition: "構造的な修正必要"
next: plan
- name: ai_review
agent: ~/.takt/agents/default/ai-antipattern-reviewer.md
allowed_tools:
- Read
- Glob
- Grep
- Write
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: ai_review (AI生成コードレビュー)
- Report Directory: .takt/reports/{report_dir}/
- Report File: .takt/reports/{report_dir}/04-ai-review.md
## Original User Request (ワークフロー開始時の元の要求)
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
AI特有の問題についてコードをレビューしてください:
- 仮定の検証
- もっともらしいが間違っているパターン
- 既存コードベースとの適合性
- スコープクリープの検出
**レポート出力:** 上記の `Report File` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**レポートフォーマット:**
```markdown
# AI生成コードレビュー
## 結果: APPROVE / REJECT
## サマリー
{1文で結果を要約}
## 検証した項目
| 観点 | 結果 | 備考 |
|------|------|------|
| 仮定の妥当性 | ✅ | - |
| API/ライブラリの実在 | ✅ | - |
| コンテキスト適合 | ✅ | - |
| スコープ | ✅ | - |
## 問題点（REJECTの場合）
| # | カテゴリ | 場所 | 問題 |
|---|---------|------|------|
| 1 | 幻覚API | `src/file.ts:23` | 存在しないメソッド |
```
**認知負荷軽減ルール:**
- 問題なし → サマリー1文 + チェック表のみ（10行以内）
- 問題あり → + 問題を表形式で（25行以内）
rules:
- condition: "AI特有の問題なし"
next: supervise
- condition: "AI特有の問題あり"
- condition: "構造的な修正必要"
next: plan
- name: supervise
agent: ~/.takt/agents/default/supervisor.md
report:
- Validation: 05-supervisor-validation.md
- Summary: summary.md
allowed_tools:
- Read
- Glob
@ -313,24 +257,11 @@ steps:
- WebSearch
- WebFetch
instruction_template: |
## Workflow Context
- Iteration: {iteration}/{max_iterations}（ワークフロー全体）
- Step Iteration: {step_iteration}（このステップの実行回数）
- Step: supervise (final verification)
- Report Directory: .takt/reports/{report_dir}/
- Report Files:
- Validation: .takt/reports/{report_dir}/05-supervisor-validation.md
- Summary: .takt/reports/{report_dir}/summary.md
## Original User Request
{task}
## Git Diff
```diff
{git_diff}
```
## Instructions
テスト実行、ビルド確認、最終承認を行ってください。
**ワークフロー全体の確認:**
@ -341,10 +272,6 @@ steps:
**レポートの確認:** Report Directory内の全レポートを読み、
未対応の改善提案がないか確認してください。
**レポート出力:** 上記の `Report Files` に出力してください。
- ファイルが存在しない場合: 新規作成
- ファイルが存在する場合: `## Iteration {step_iteration}` セクションを追記
**Validationレポートフォーマット:**
```markdown
# 最終検証結果
@ -387,8 +314,8 @@ steps:
## レビュー結果
| レビュー | 結果 |
|---------|------|
| Architect | ✅ APPROVE |
| AI Review | ✅ APPROVE |
| Architect | ✅ APPROVE |
| Supervisor | ✅ APPROVE |
## 確認コマンド

199
src/__tests__/instructionBuilder.test.ts
View File

@ -344,6 +344,205 @@ describe('instruction-builder', () => {
});
});
describe('auto-injected Workflow Context section', () => {
it('should include iteration, step iteration, and step name', () => {
const step = createMinimalStep('Do work');
step.name = 'implement';
const context = createMinimalContext({
iteration: 3,
maxIterations: 20,
stepIteration: 2,
language: 'en',
});
const result = buildInstruction(step, context);
expect(result).toContain('## Workflow Context');
expect(result).toContain('- Iteration: 3/20');
expect(result).toContain('- Step Iteration: 2');
expect(result).toContain('- Step: implement');
});
it('should include single report file when report is a string', () => {
const step = createMinimalStep('Do work');
step.name = 'plan';
step.report = '00-plan.md';
const context = createMinimalContext({
reportDir: '20260129-test',
language: 'en',
});
const result = buildInstruction(step, context);
expect(result).toContain('- Report Directory: .takt/reports/20260129-test/');
expect(result).toContain('- Report File: .takt/reports/20260129-test/00-plan.md');
expect(result).not.toContain('Report Files:');
});
it('should include multiple report files when report is ReportConfig[]', () => {
const step = createMinimalStep('Do work');
step.name = 'implement';
step.report = [
{ label: 'Scope', path: '01-scope.md' },
{ label: 'Decisions', path: '02-decisions.md' },
];
const context = createMinimalContext({
reportDir: '20260129-test',
language: 'en',
});
const result = buildInstruction(step, context);
expect(result).toContain('- Report Directory: .takt/reports/20260129-test/');
expect(result).toContain('- Report Files:');
expect(result).toContain(' - Scope: .takt/reports/20260129-test/01-scope.md');
expect(result).toContain(' - Decisions: .takt/reports/20260129-test/02-decisions.md');
expect(result).not.toContain('Report File:');
});
it('should NOT include report info when reportDir is undefined', () => {
const step = createMinimalStep('Do work');
step.report = '00-plan.md';
const context = createMinimalContext({ language: 'en' });
const result = buildInstruction(step, context);
expect(result).toContain('## Workflow Context');
expect(result).not.toContain('Report Directory');
expect(result).not.toContain('Report File');
});
it('should NOT include report info when step has no report', () => {
const step = createMinimalStep('Do work');
const context = createMinimalContext({
reportDir: '20260129-test',
language: 'en',
});
const result = buildInstruction(step, context);
expect(result).toContain('## Workflow Context');
expect(result).not.toContain('Report Directory');
expect(result).not.toContain('Report File');
});
it('should render Japanese step iteration suffix', () => {
const step = createMinimalStep('Do work');
const context = createMinimalContext({
stepIteration: 3,
language: 'ja',
});
const result = buildInstruction(step, context);
expect(result).toContain('- Step Iteration: 3（このステップの実行回数）');
});
});
describe('auto-injected User Request and Additional User Inputs sections', () => {
it('should include User Request section with task', () => {
const step = createMinimalStep('Do work');
const context = createMinimalContext({ task: 'Build the feature', language: 'en' });
const result = buildInstruction(step, context);
expect(result).toContain('## User Request\n');
});
it('should include Additional User Inputs section', () => {
const step = createMinimalStep('Do work');
const context = createMinimalContext({
userInputs: ['input1', 'input2'],
language: 'en',
});
const result = buildInstruction(step, context);
expect(result).toContain('## Additional User Inputs\n');
});
it('should include Previous Response when passPreviousResponse is true and output exists', () => {
const step = createMinimalStep('Do work');
step.passPreviousResponse = true;
const context = createMinimalContext({
previousOutput: { content: 'Previous result', tag: '[TEST:1]' },
language: 'en',
});
const result = buildInstruction(step, context);
expect(result).toContain('## Previous Response\n');
});
it('should NOT include Previous Response when passPreviousResponse is false', () => {
const step = createMinimalStep('Do work');
step.passPreviousResponse = false;
const context = createMinimalContext({
previousOutput: { content: 'Previous result', tag: '[TEST:1]' },
language: 'en',
});
const result = buildInstruction(step, context);
expect(result).not.toContain('## Previous Response');
});
it('should include Instructions header before template content', () => {
const step = createMinimalStep('My specific instructions here');
const context = createMinimalContext({ language: 'en' });
const result = buildInstruction(step, context);
const instructionsIdx = result.indexOf('## Instructions');
const contentIdx = result.indexOf('My specific instructions here');
expect(instructionsIdx).toBeGreaterThan(-1);
expect(contentIdx).toBeGreaterThan(instructionsIdx);
});
it('should skip auto-injected User Request when template contains {task}', () => {
const step = createMinimalStep('Process this: {task}');
const context = createMinimalContext({ task: 'My task', language: 'en' });
const result = buildInstruction(step, context);
// Auto-injected section should NOT appear
expect(result).not.toContain('## User Request');
// But template placeholder should be replaced
expect(result).toContain('Process this: My task');
});
it('should skip auto-injected Previous Response when template contains {previous_response}', () => {
const step = createMinimalStep('## Feedback\n{previous_response}\n\nFix the issues.');
step.passPreviousResponse = true;
const context = createMinimalContext({
previousOutput: { content: 'Review feedback here', tag: '[TEST:1]' },
language: 'en',
});
const result = buildInstruction(step, context);
// Auto-injected section should NOT appear
expect(result).not.toContain('## Previous Response\n');
// But template placeholder should be replaced with content
expect(result).toContain('## Feedback\nReview feedback here');
});
it('should skip auto-injected Additional User Inputs when template contains {user_inputs}', () => {
const step = createMinimalStep('Inputs: {user_inputs}');
const context = createMinimalContext({
userInputs: ['extra info'],
language: 'en',
});
const result = buildInstruction(step, context);
// Auto-injected section should NOT appear
expect(result).not.toContain('## Additional User Inputs');
// But template placeholder should be replaced
expect(result).toContain('Inputs: extra info');
});
});
describe('basic placeholder replacement', () => {
it('should replace {task} placeholder', () => {
const step = createMinimalStep('Execute: {task}');

25
src/config/workflowLoader.ts
View File

@ -8,7 +8,7 @@ import { readFileSync, existsSync, readdirSync, statSync } from 'node:fs';
import { join, dirname, basename } from 'node:path';
import { parse as parseYaml } from 'yaml';
import { WorkflowConfigRawSchema } from '../models/schemas.js';
import type { WorkflowConfig, WorkflowStep, WorkflowRule } from '../models/types.js';
import type { WorkflowConfig, WorkflowStep, WorkflowRule, ReportConfig } from '../models/types.js';
import { getGlobalWorkflowsDir } from './paths.js';
/** Get builtin workflow by name */
@ -54,6 +54,28 @@ function extractAgentDisplayName(agentPath: string): string {
return filename;
}
/**
* Normalize the raw report field from YAML into internal format.
*
* YAML formats:
* report: "00-plan.md" → string (single file)
* report: → ReportConfig[] (multiple files)
* - Scope: 01-scope.md
* - Decisions: 02-decisions.md
*
* Array items are parsed as single-key objects: [{Scope: "01-scope.md"}, ...]
*/
function normalizeReport(
raw: string | Record<string, string>[] | undefined,
): string | ReportConfig[] | undefined {
if (raw == null) return undefined;
if (typeof raw === 'string') return raw;
// Convert [{Scope: "01-scope.md"}, ...] to [{label: "Scope", path: "01-scope.md"}, ...]
return raw.flatMap((entry) =>
Object.entries(entry).map(([label, path]) => ({ label, path })),
);
}
/**
* Convert raw YAML workflow config to internal format.
* Agent paths are resolved relative to the workflow directory.
@ -79,6 +101,7 @@ function normalizeWorkflowConfig(raw: unknown, workflowDir: string): WorkflowCon
permissionMode: step.permission_mode,
instructionTemplate: step.instruction_template || step.instruction || '{task}',
rules,
report: normalizeReport(step.report),
passPreviousResponse: step.pass_previous_response,
};
});

1
src/models/index.ts
View File

@ -2,6 +2,7 @@
export type {
AgentType,
Status,
ReportConfig,
AgentResponse,
SessionState,
WorkflowStep,

18
src/models/schemas.ts
View File

@ -27,6 +27,22 @@ export const StatusSchema = z.enum([
/** Permission mode schema for tool execution */
export const PermissionModeSchema = z.enum(['default', 'acceptEdits', 'bypassPermissions']);
/**
* Report field schema.
*
* YAML formats:
* report: 00-plan.md # single file
* report: # multiple files (label: path map entries)
* - Scope: 01-scope.md
* - Decisions: 02-decisions.md
*
* Array items are parsed as single-key objects: [{Scope: "01-scope.md"}, ...]
*/
export const ReportFieldSchema = z.union([
z.string().min(1),
z.array(z.record(z.string(), z.string())).min(1),
]);
/** Rule-based transition schema (new unified format) */
export const WorkflowRuleSchema = z.object({
/** Human-readable condition text */
@ -52,6 +68,8 @@ export const WorkflowStepRawSchema = z.object({
instruction_template: z.string().optional(),
/** Rules for step routing */
rules: z.array(WorkflowRuleSchema).optional(),
/** Report file(s) for this step */
report: ReportFieldSchema.optional(),
pass_previous_response: z.boolean().optional().default(true),
});

10
src/models/types.ts
View File

@ -50,6 +50,14 @@ export interface WorkflowRule {
appendix?: string;
}
/** Report file configuration for a workflow step */
export interface ReportConfig {
/** Display label (e.g., "Scope", "Decisions") */
label: string;
/** File path relative to report directory (e.g., "01-coder-scope.md") */
path: string;
}
/** Permission mode for tool execution */
export type PermissionMode = 'default' | 'acceptEdits' | 'bypassPermissions';
@ -73,6 +81,8 @@ export interface WorkflowStep {
instructionTemplate: string;
/** Rules for step routing */
rules?: WorkflowRule[];
/** Report file configuration. Single string for one file, array for multiple. */
report?: string | ReportConfig[];
passPreviousResponse: boolean;
}

214
src/workflow/instruction-builder.ts
View File

@ -1,11 +1,14 @@
/**
* Instruction template builder for workflow steps
*
* Builds the instruction string for agent execution by replacing
* template placeholders with actual values.
* Builds the instruction string for agent execution by:
* 1. Auto-injecting standard sections (Execution Context, Workflow Context,
* User Request, Previous Response, Additional User Inputs, Instructions header)
* 2. Replacing template placeholders with actual values
* 3. Appending auto-generated status rules from workflow rules
*/
import type { WorkflowStep, WorkflowRule, AgentResponse, Language } from '../models/types.js';
import type { WorkflowStep, WorkflowRule, AgentResponse, Language, ReportConfig } from '../models/types.js';
import { getGitDiff } from '../agents/runner.js';
/**
@ -213,76 +216,207 @@ function escapeTemplateChars(str: string): string {
return str.replace(/\{/g, '｛').replace(/\}/g, '｝');
}
/** Localized strings for auto-injected sections */
const SECTION_STRINGS = {
en: {
workflowContext: '## Workflow Context',
iteration: 'Iteration',
iterationWorkflowWide: '(workflow-wide)',
stepIteration: 'Step Iteration',
stepIterationTimes: '(times this step has run)',
step: 'Step',
reportDirectory: 'Report Directory',
reportFile: 'Report File',
reportFiles: 'Report Files',
userRequest: '## User Request',
previousResponse: '## Previous Response',
additionalUserInputs: '## Additional User Inputs',
instructions: '## Instructions',
},
ja: {
workflowContext: '## Workflow Context',
iteration: 'Iteration',
iterationWorkflowWide: '（ワークフロー全体）',
stepIteration: 'Step Iteration',
stepIterationTimes: '（このステップの実行回数）',
step: 'Step',
reportDirectory: 'Report Directory',
reportFile: 'Report File',
reportFiles: 'Report Files',
userRequest: '## User Request',
previousResponse: '## Previous Response',
additionalUserInputs: '## Additional User Inputs',
instructions: '## Instructions',
},
} as const;
/**
* Build instruction from template with context values.
*
* Supported placeholders:
* - {task} - The main task/prompt
* - {iteration} - Current iteration number (workflow-wide turn count)
* - {max_iterations} - Maximum iterations allowed
* - {step_iteration} - Current step's iteration number (how many times this step has been executed)
* - {previous_response} - Output from previous step (if passPreviousResponse is true)
* - {git_diff} - Current git diff output
* - {user_inputs} - Accumulated user inputs
* - {report_dir} - Report directory name (e.g., "20250126-143052-task-summary")
* Render the Workflow Context section.
*/
export function buildInstruction(
function renderWorkflowContext(
step: WorkflowStep,
context: InstructionContext
context: InstructionContext,
language: Language,
): string {
let instruction = step.instructionTemplate;
const s = SECTION_STRINGS[language];
const lines: string[] = [
s.workflowContext,
`- ${s.iteration}: ${context.iteration}/${context.maxIterations}${s.iterationWorkflowWide}`,
`- ${s.stepIteration}: ${context.stepIteration}${s.stepIterationTimes}`,
`- ${s.step}: ${step.name}`,
];
// Report info (only if step has report config AND reportDir is available)
if (step.report && context.reportDir) {
lines.push(`- ${s.reportDirectory}: .takt/reports/${context.reportDir}/`);
if (typeof step.report === 'string') {
// Single file
lines.push(`- ${s.reportFile}: .takt/reports/${context.reportDir}/${step.report}`);
} else {
// Multiple files
lines.push(`- ${s.reportFiles}:`);
for (const file of step.report as ReportConfig[]) {
lines.push(` - ${file.label}: .takt/reports/${context.reportDir}/${file.path}`);
}
}
}
return lines.join('\n');
}
/**
* Replace template placeholders in the instruction_template body.
*
* These placeholders may still be used in instruction_template for
* backward compatibility or special cases (e.g., {git_diff} in review steps).
*/
function replaceTemplatePlaceholders(
template: string,
step: WorkflowStep,
context: InstructionContext,
): string {
let result = template;
// These placeholders are also covered by auto-injected sections
// (User Request, Previous Response, Additional User Inputs), but kept here
// for backward compatibility with workflows that still embed them in
// instruction_template (e.g., research.yaml, magi.yaml).
// New workflows should NOT use {task} or {user_inputs} in instruction_template
// since they are auto-injected as separate sections.
// Replace {task}
instruction = instruction.replace(/\{task\}/g, escapeTemplateChars(context.task));
result = result.replace(/\{task\}/g, escapeTemplateChars(context.task));
// Replace {iteration}, {max_iterations}, and {step_iteration}
instruction = instruction.replace(/\{iteration\}/g, String(context.iteration));
instruction = instruction.replace(/\{max_iterations\}/g, String(context.maxIterations));
instruction = instruction.replace(/\{step_iteration\}/g, String(context.stepIteration));
result = result.replace(/\{iteration\}/g, String(context.iteration));
result = result.replace(/\{max_iterations\}/g, String(context.maxIterations));
result = result.replace(/\{step_iteration\}/g, String(context.stepIteration));
// Replace {previous_response}
if (step.passPreviousResponse) {
if (context.previousOutput) {
instruction = instruction.replace(
result = result.replace(
/\{previous_response\}/g,
escapeTemplateChars(context.previousOutput.content)
escapeTemplateChars(context.previousOutput.content),
);
} else {
instruction = instruction.replace(/\{previous_response\}/g, '');
result = result.replace(/\{previous_response\}/g, '');
}
}
// Replace {git_diff}
const gitDiff = getGitDiff(context.cwd);
instruction = instruction.replace(/\{git_diff\}/g, gitDiff);
result = result.replace(/\{git_diff\}/g, gitDiff);
// Replace {user_inputs}
const userInputsStr = context.userInputs.join('\n');
instruction = instruction.replace(
result = result.replace(
/\{user_inputs\}/g,
escapeTemplateChars(userInputsStr)
escapeTemplateChars(userInputsStr),
);
// Replace {report_dir} with the directory name, keeping paths relative.
// In worktree mode, a symlink from cwd/.takt/reports → projectCwd/.takt/reports
// ensures the relative path resolves correctly without embedding absolute paths
// that could cause agents to operate on the wrong repository.
// Replace {report_dir}
if (context.reportDir) {
instruction = instruction.replace(/\{report_dir\}/g, context.reportDir);
result = result.replace(/\{report_dir\}/g, context.reportDir);
}
// Append auto-generated status rules from rules
return result;
}
/**
* Build instruction from template with context values.
*
* Generates a complete instruction by auto-injecting standard sections
* around the step-specific instruction_template content:
*
* 1. Execution Context (working directory, rules) — always
* 2. Workflow Context (iteration, step, report info) — always
* 3. User Request ({task}) — unless template contains {task}
* 4. Previous Response — if passPreviousResponse and has content, unless template contains {previous_response}
* 5. Additional User Inputs — unless template contains {user_inputs}
* 6. Instructions header + instruction_template content — always
* 7. Status Output Rules — if rules exist
*
* Template placeholders ({task}, {git_diff}, etc.) are still replaced
* within the instruction_template body for backward compatibility.
* When a placeholder is present in the template, the corresponding
* auto-injected section is skipped to avoid duplication.
*/
export function buildInstruction(
step: WorkflowStep,
context: InstructionContext,
): string {
const language = context.language ?? 'en';
const s = SECTION_STRINGS[language];
const sections: string[] = [];
// 1. Execution context metadata (working directory + rules)
const metadata = buildExecutionMetadata(context);
sections.push(renderExecutionMetadata(metadata));
// 2. Workflow Context (iteration, step, report info)
sections.push(renderWorkflowContext(step, context, language));
// Skip auto-injection for sections whose placeholders exist in the template,
// to avoid duplicate content. Templates using placeholders handle their own layout.
const tmpl = step.instructionTemplate;
const hasTaskPlaceholder = tmpl.includes('{task}');
const hasPreviousResponsePlaceholder = tmpl.includes('{previous_response}');
const hasUserInputsPlaceholder = tmpl.includes('{user_inputs}');
// 3. User Request (skip if template embeds {task} directly)
if (!hasTaskPlaceholder) {
sections.push(`${s.userRequest}\n${escapeTemplateChars(context.task)}`);
}
// 4. Previous Response (skip if template embeds {previous_response} directly)
if (step.passPreviousResponse && context.previousOutput && !hasPreviousResponsePlaceholder) {
sections.push(
`${s.previousResponse}\n${escapeTemplateChars(context.previousOutput.content)}`,
);
}
// 5. Additional User Inputs (skip if template embeds {user_inputs} directly)
if (!hasUserInputsPlaceholder) {
const userInputsStr = context.userInputs.join('\n');
sections.push(`${s.additionalUserInputs}\n${escapeTemplateChars(userInputsStr)}`);
}
// 6. Instructions header + instruction_template content
const processedTemplate = replaceTemplatePlaceholders(
step.instructionTemplate,
step,
context,
);
sections.push(`${s.instructions}\n${processedTemplate}`);
// 7. Status rules (auto-generated from rules)
if (step.rules && step.rules.length > 0) {
const statusHeader = renderStatusRulesHeader(language);
const generatedPrompt = generateStatusRulesFromRules(step.name, step.rules, language);
instruction = `${instruction}\n\n${statusHeader}\n${generatedPrompt}`;
sections.push(`${statusHeader}\n${generatedPrompt}`);
}
// Prepend execution context metadata so agents see it first.
// Now language-aware, so no need to hide it at the end.
const metadata = buildExecutionMetadata(context);
instruction = `${renderExecutionMetadata(metadata)}\n${instruction}`;
return instruction;
return sections.join('\n\n');
}