doc-tasks-autopilot（文档任务自动驾驶仪）

doc-tasks-autopilot

Purpose

Automated Task Breakdown (TASKS) generation pipeline that processes SPEC and TSPEC documents to generate AI-structured TODO tasks for implementation with CODE-Ready scoring.

Layer: 11 (Final documentation layer before code)

Upstream: SPEC (Layer 9), TSPEC (Layer 10)

Downstream: Code (Layer 12 - execution layer)

Input Contract (IPLAN-004 Standard)

• Supported modes:

- --ref <path> - --prompt "<text>" - --iplan <path|IPLAN-NNN>

• Precedence: --iplan > --ref > --prompt
• IPLAN resolution order:

1. Use explicit file path when it exists 2. Resolve work_plans/IPLAN-NNN*.md 3. Resolve governance/plans/IPLAN-NNN*.md 4. If multiple matches exist, fail with disambiguation request

• Merge conflict rule:

- Objective/scope conflicts between primary and supplemental sources are blocking and require user clarification.

Input Contract (IPLAN-004 Standard)

• Supported modes:

- --ref <path> - --prompt "<text>" - --iplan <path|IPLAN-NNN>

• Precedence: --iplan > --ref > --prompt
• IPLAN resolution order:

1. Use explicit file path when it exists 2. Resolve work_plans/IPLAN-NNN*.md 3. Resolve governance/plans/IPLAN-NNN*.md 4. If multiple matches exist, fail with disambiguation request

• Merge conflict rule:

- Objective/scope conflicts between primary and supplemental sources are blocking and require user clarification.

Skill Dependencies

Document Type Contract (MANDATORY)

When generating TASKS document instances, the autopilot MUST:

1. Read instance_document_type from template:

- Source: ai_dev_ssd_flow/11_TASKS/TASKS-MVP-TEMPLATE.yaml - Field: metadata.instance_document_type: "tasks-document"

1. Set document_type in generated document frontmatter: custom_fields: document_type: tasks-document # NOT "template" artifact_type: TASKS layer: 11
2. Validation: Generated documents MUST have document_type: tasks-document

- Templates have document_type: template - Instances have document_type: tasks-document - Schema validates both values

Error Handling: If instance_document_type is missing from template, default to tasks-document.

Smart Document Detection

The autopilot automatically determines the action based on the input document type.

Input Type Recognition (Multiple Upstreams)

TASKS can be derived from SPEC and/or TSPEC:

Detection Algorithm

1. Parse input: Extract TYPE and NN from "{TYPE}-{NN}"
2. Determine action:
   - IF TYPE == "TASKS": Review Mode
   - ELSE IF TYPE in ["SPEC", "TSPEC"]: Generate/Find Mode
   - ELSE: Error (invalid type for this autopilot)
3. For Generate/Find Mode:
   - Check: Does TASKS-{NN} exist in docs/11_TASKS/?
   - IF exists: Switch to Review Mode for TASKS-{NN}
   - ELSE: Proceed with Generation from {TYPE}-{NN}

File Existence Check

# Check for nested folder structure (mandatory)
ls docs/11_TASKS/TASKS-{NN}_*/

Examples

# Review mode (same type - TASKS input)
/doc-tasks-autopilot TASKS-01         # Reviews existing TASKS-01

# Generate/Find mode (upstream types)
/doc-tasks-autopilot SPEC-01          # Generates TASKS-01 if missing, or reviews existing TASKS-01
/doc-tasks-autopilot TSPEC-01         # Generates TASKS-01 if missing, or reviews existing TASKS-01

# Multiple inputs
/doc-tasks-autopilot SPEC-01,SPEC-02  # Generates/reviews TASKS-01 and TASKS-02
/doc-tasks-autopilot TASKS-01,TASKS-02 # Reviews TASKS-01 and TASKS-02

Action Determination Output

Input: SPEC-01
├── Detected Type: SPEC (primary upstream)
├── Expected TASKS: TASKS-01
├── TASKS Exists: Yes → docs/11_TASKS/TASKS-01_f1_iam/
└── Action: REVIEW MODE - Running doc-tasks-reviewer on TASKS-01

Input: TSPEC-05
├── Detected Type: TSPEC (alternative upstream)
├── Expected TASKS: TASKS-05
├── TASKS Exists: No
└── Action: GENERATE MODE - Creating TASKS-05 from TSPEC-05

Input: TASKS-03
├── Detected Type: TASKS (self)
└── Action: REVIEW MODE - Running doc-tasks-reviewer on TASKS-03

Workflow Overview

flowchart TD
    subgraph Phase1["Phase 1: SPEC/TSPEC Analysis"]
        A[Start] --> B[Read SPEC Documents]
        B --> C[Read TSPEC Documents]
        C --> D[Extract Implementation Units]
        D --> E[Extract Test Requirements]
        E --> F[Catalog Work Items]
    end

    subgraph Phase2["Phase 2: Task Decomposition Planning"]
        F --> G[Check SPEC TASKS-Ready Score]
        G --> H[Check TSPEC TASKS-Ready Score]
        H --> I{Both >= 90%?}
        I -->|No| J[Flag Issues]
        J --> K{Auto-Fixable?}
        K -->|Yes| L[Fix Issues]
        L --> G
        K -->|No| M[Abort - Manual Fix Required]
        I -->|Yes| N[Plan Task Decomposition]
        N --> O[Identify Dependencies]
    end

    subgraph Phase3["Phase 3: TASKS Generation"]
        O --> P[Generate Implementation Tasks]
        P --> Q[Generate Test Tasks]
        Q --> R[Build Dependency Graph]
        R --> S[Assign Priorities]
        S --> T[quality-advisor: Real-time Feedback]
        T --> U[Generate Implementation Contracts]
        U --> V[Add Traceability Tags]
        V --> W[Write TASKS File]
    end

    subgraph Phase4["Phase 4: TASKS Validation"]
        W --> X[Run doc-tasks-validator]
        X --> Y{CODE-Ready >= 90%?}
        Y -->|No| Z[Auto-Fix TASKS Issues]
        Z --> AA[Re-validate TASKS]
        AA --> Y
        Y -->|Yes| AB[Mark TASKS Validated]
    end

    subgraph Phase5["Phase 5: Review & Fix Cycle"]
        AB --> AC[Run doc-tasks-reviewer]
        AC --> AC2{Score >= 90?}
        AC2 -->|No| AC3[Run doc-tasks-fixer]
        AC3 --> AC4{Iteration < Max?}
        AC4 -->|Yes| AC
        AC4 -->|No| AC5[Flag Manual Review]
        AC2 -->|Yes| AD[Verify Quality Checks]
        AC5 --> AD
        AD --> AE[Update Traceability Matrix]
        AE --> AF[Generate Summary Report]
    end

    AF --> AG[Complete]
    M --> AI[Exit with Error]

Task Structure

Element ID Format

Priority Levels

TASKS Document Structure

# TASKS-01: Component Implementation

## Document Control
| Field | Value |
|-------|-------|
| Status | Draft |
| CODE-Ready Score | 95% |

## 1. Task Overview
Summary of implementation tasks derived from SPEC-01.

## 2. Dependency Graph

graph TD T1[TASKS.01.1801] --> T2[TASKS.01.1802] T1 --> T3[TASKS.01.1803] T2 --> T4[TASKS.01.1804] T3 --> T4

## 3. Implementation Tasks

### TASKS.01.1801: Setup Project Structure

- **Priority**: P0
- **Upstream**: None
- **Downstream**: TASKS.01.1802, TASKS.01.1803
- **SPEC Reference**: SPEC.01.2801
- **Estimated Complexity**: 2/5

### TASKS.01.1802: Implement Core Logic

- **Priority**: P1
- **Upstream**: TASKS.01.1801
- **Downstream**: TASKS.01.1804
- **SPEC Reference**: SPEC.01.2802

## 7. Implementation Contracts

[Protocol interfaces, exception hierarchies, state machine contracts]

## 8. Traceability

@brd: BRD.01.0101 @prd: PRD.01.0901...

Implementation Contracts

Per IMPLEMENTATION_CONTRACTS_GUIDE.md, generate contracts when:

• TASKS file has 3+ downstream dependencies
• Shared interfaces across multiple sessions
• Complex state machines or exception hierarchies

Contract Types:

1. Protocol Interfaces
2. Exception Hierarchies
3. State Machine Contracts
4. Data Models
5. Dependency Injection Interfaces

Phase 5: Review & Fix Cycle (v2.3)

Iterative review and fix cycle to ensure TASKS quality before completion.

flowchart TD
    A[Phase 5 Start] --> B[Run doc-tasks-reviewer]
    B --> C[Generate Review Report]
    C --> D{Review Score >= 90?}

    D -->|Yes| E[PASS - Proceed to Phase 6]
    D -->|No| F{Iteration < Max?}

    F -->|Yes| G[Run doc-tasks-fixer]
    G --> H[Apply Fixes]
    H --> I[Generate Fix Report]
    I --> J[Increment Iteration]
    J --> B

    F -->|No| K[Flag for Manual Review]
    K --> L[Generate Final Report with Remaining Issues]
    L --> E

5.1 Initial Review

Run doc-tasks-reviewer to identify issues.

/doc-tasks-reviewer TASKS-NN

Output: TASKS-NN.R_review_report_v001.md

5.2 Fix Cycle

If review score < 90%, invoke doc-tasks-fixer.

/doc-tasks-fixer TASKS-NN --revalidate

Fix Categories:

Output: TASKS-NN.F_fix_report_v001.md

5.3 Re-Review

After fixes, automatically re-run reviewer.

/doc-tasks-reviewer TASKS-NN

Output: TASKS-NN.R_review_report_v002.md

5.4 Iteration Control

Iteration Example:

Iteration 1:
  Review v001: Score 75 (3 errors, 5 warnings)
  Fix v001: Fixed 6 issues, regenerated dependency graph

Iteration 2:
  Review v002: Score 92 (0 errors, 2 warnings)
  Status: PASS (score >= 90)

5.5 Quality Checks (Post-Fix)

After passing the fix cycle:

1. Task Structure Completeness:

- All required task fields present (Priority, Upstream, Downstream) - SPEC references valid and linked - Complexity estimates provided

1. Dependency Graph Validity:

- Valid DAG (no circular dependencies) - All task references resolved - Mermaid diagram renders correctly

1. Element ID Compliance (per doc-naming skill):

- All IDs use TASKS.NN.xxxx format - Element type codes valid for TASKS (18, 30) - No legacy patterns (TASK-XXX, TODO-XXX, TI-XXX)

1. CODE-Ready Report: CODE-Ready Score Breakdown ========================== Task Completeness: 24/25 (all required fields) Dependency Graph: 20/20 (valid DAG) Priority Assignments: 15/15 (P0-P3 assigned) Implementation Contracts: 14/15 (contracts where needed) Traceability Tags: 15/15 (all 9 tags present) SPEC/TSPEC Alignment: 10/10 (references valid) ---------------------------- Total CODE-Ready Score: 98/100 (Target: >= 90) Status: READY FOR CODE IMPLEMENTATION
2. Traceability Matrix Update: # Update TASKS traceability python ai_dev_flow/scripts/update_traceability_matrix.py \ --tasks docs/11_TASKS/TASKS-NN_{slug}.md \ --matrix docs/11_TASKS/TASKS-00_TRACEABILITY_MATRIX.md

Cumulative Tags (9 Required)

@brd: BRD.NN.xxxx
@prd: PRD.NN.xxxx
@ears: EARS.NN.xxxx
@bdd: BDD.NN.xxxx
@adr: ADR-NN
@sys: SYS.NN.xxxx
@req: REQ.NN.xxxx
@spec: SPEC.NN.xxxx
@tspec: TSPEC.NN.xxxx
@ctr: CTR.NN.xxxx  # Optional

Configuration

Default Configuration

tasks_autopilot:
  version: "1.0"

  scoring:
    tasks_ready_min: 90
    code_ready_min: 90
    strict_mode: false

  execution:
    max_parallel: 3        # HARD LIMIT - do not exceed
    chunk_size: 3          # Documents per chunk
    pause_between_chunks: true
    auto_fix: true
    continue_on_error: false
    timeout_per_spec: 180  # seconds

  output:
    report_format: markdown

  validation:
    skip_validation: false
    fix_iterations_max: 3

  contracts:
    generate_when_dependencies_gte: 3
    include_protocol_interfaces: true
    include_exception_hierarchies: true
    include_state_machines: true

Execution Modes

Mode 1: Generate Mode (Default)

Standard TASKS generation from SPEC/TSPEC documents (see Workflow Overview above).

Mode 2: Review Mode (v2.1)

Validate existing TASKS documents without modification. Generates quality report with actionable recommendations.

Command:

# Review single TASKS
/doc-tasks-autopilot TASKS-01 --review

# Review all TASKS in directory
/doc-tasks-autopilot docs/11_TASKS/ --review --all

# Review with detailed report
/doc-tasks-autopilot TASKS-01 --review --verbose

Review Process:

flowchart TD
    A[Input: TASKS Document] --> B[Load TASKS Content]
    B --> C[Validate Task Structure]
    C --> D[Check CODE-Ready Score]
    D --> E[Verify Dependency Graph]
    E --> F[Validate Element IDs]
    F --> G[Check Implementation Contracts]
    G --> H[Verify Cumulative Tags]
    H --> I[Generate Review Report]
    I --> J{Issues Found?}
    J -->|Yes| K[Categorize Issues]
    K --> L[Generate Fix Recommendations]
    L --> M[Output Review Report]
    J -->|No| M

Review Report Template:

# TASKS Review Report: TASKS-NN_{slug}

## Summary
- **CODE-Ready Score**: NN% (✅/🟡/❌)
- **Total Issues**: N (E errors, W warnings)
- **Auto-Fixable**: N issues
- **Manual Review**: N issues

## Score Breakdown
| Category | Score | Max | Status |
|----------|-------|-----|--------|
| Task Completeness | NN | 25 | ✅/🟡/❌ |
| Dependency Graph | NN | 20 | ✅/🟡/❌ |
| Priority Assignments | NN | 15 | ✅/🟡/❌ |
| Implementation Contracts | NN | 15 | ✅/🟡/❌ |
| Traceability Tags | NN | 15 | ✅/🟡/❌ |
| SPEC/TSPEC Alignment | NN | 10 | ✅/🟡/❌ |

## Issues by Category

### Auto-Fixable Issues
| Issue | Location | Fix Action |
|-------|----------|------------|
| Legacy ID pattern | Line 45 | Convert TASK-001 → TASKS.01.1801 |
| Missing cumulative tag | Traceability | Add @sys: SYS.01.0101 |

### Manual Review Required
| Issue | Location | Recommendation |
|-------|----------|----------------|
| Circular dependency | T3 ↔ T5 | Restructure task dependencies |
| Missing complexity | TASKS.01.1804 | Add Estimated Complexity |

Score Indicators:

• ✅ Green (>=90%): CODE-Ready
• 🟡 Yellow (70-89%): Needs improvement
• ❌ Red (<70%): Significant issues

Review Configuration:

review_mode:
  enabled: true
  checks:
    - task_structure       # All required task fields present
    - dependency_graph     # Valid DAG, no cycles
    - priority_assignments # P0-P3 properly assigned
    - implementation_contracts # Contracts when 3+ dependencies
    - cumulative_tags      # 9 required tags present
    - element_ids          # TASKS.NN.18.SS, TASKS.NN.30.SS format
    - spec_alignment       # Tasks trace to SPEC/TSPEC
  output:
    format: markdown       # markdown, json, html
    include_recommendations: true
    include_fix_commands: true

Mode 3: Fix Mode (v2.1)

Auto-repair existing TASKS documents with backup and content preservation.

Command:

# Fix single TASKS
/doc-tasks-autopilot TASKS-01 --fix

# Fix with backup
/doc-tasks-autopilot TASKS-01 --fix --backup

# Fix all TASKS
/doc-tasks-autopilot docs/11_TASKS/ --fix --all

# Fix specific categories only
/doc-tasks-autopilot TASKS-01 --fix --only element_ids,tags

# Dry-run fix (preview changes)
/doc-tasks-autopilot TASKS-01 --fix --dry-run

Fix Process:

flowchart TD
    A[Input: TASKS Document] --> B[Create Backup]
    B --> C[Load TASKS Content]
    C --> D[Run Review Analysis]
    D --> E{Auto-Fixable Issues?}
    E -->|No| F[Report: No Fixes Needed]
    E -->|Yes| G[Apply Fixes by Category]

    subgraph FixCategories["Fix Categories"]
        G --> H[Fix Element IDs]
        H --> I[Fix Priority Format]
        I --> J[Regenerate Dependency Graph]
        J --> K[Add Missing Tags]
        K --> L[Fix Section Structure]
        L --> M[Add Implementation Contracts]
    end

    M --> N[Validate Fixed Document]
    N --> O{Validation Passed?}
    O -->|No| P[Rollback to Backup]
    P --> Q[Report: Fix Failed]
    O -->|Yes| R[Save Fixed Document]
    R --> S[Generate Fix Report]

TASKS-Specific Fix Categories:

Element ID Migration (Layer 11):

Content Preservation Rules:

Fix Configuration:

fix_mode:
  enabled: true
  backup:
    enabled: true
    location: "tmp/backups/"
    timestamp: true
  fix_categories:
    element_ids: true       # Convert legacy ID patterns
    priorities: true        # Normalize P0-P3 format
    dependency_graph: true  # Regenerate Mermaid diagram
    cumulative_tags: true   # Add 9 required tags
    sections: true          # Add missing sections
    contracts: true         # Generate when criteria met
    circular_deps: false    # Manual only (flag but don't auto-fix)
  validation:
    post_fix: true          # Validate after fixes
    rollback_on_fail: true  # Restore backup if validation fails
  preserve:
    task_descriptions: true
    dependency_logic: true
    priority_values: true
    contract_code: true

Fix Report Template:

# TASKS Fix Report: TASKS-NN_{slug}

## Summary
- **Backup Created**: tmp/backups/TASKS-NN_{slug}_20260209_143022.md
- **Issues Fixed**: N of M auto-fixable issues
- **Manual Review**: N issues flagged

## Fixes Applied

### Element ID Migration
| Original | Fixed | Location |
|----------|-------|----------|
| TASK-001 | TASKS.01.1801 | Line 45 |
| TI-001 | TASKS.01.3001 | Line 78 |

### Dependency Graph Regenerated
- Updated Mermaid diagram with corrected task references
- Total tasks: 12, Dependencies: 18

### Cumulative Tags Added
- @ears: EARS.01.25001 (added)
- @sys: SYS.01.0101 (added)
- @req: REQ.01.0101 (added)

### Implementation Contracts Generated
- Protocol interface: `IDataProcessor` (TASKS.01.1803 has 4 dependencies)
- Exception hierarchy: `TaskExecutionError` family

## Manual Review Required

### Circular Dependencies
| Cycle | Tasks Involved | Recommendation |
|-------|----------------|----------------|
| Cycle 1 | T3 → T5 → T3 | Restructure: T5 should not depend on T3 |

## Validation Results
- **CODE-Ready Score**: Before: 72% → After: 94%
- **Validation Errors**: Before: 8 → After: 0
- **Status**: ✅ All auto-fixes validated

Command Line Options (Review/Fix Modes):

Context Management

Chunked Parallel Execution (MANDATORY)

CRITICAL: To prevent conversation context overflow errors ("Prompt is too long", "Conversation too long"), all autopilot operations MUST follow chunked execution rules:

Chunk Size Limit: Maximum 3 documents per chunk

Chunking Rules:

1. Chunk Formation: Group SPEC/TSPEC-derived TASKS documents into chunks of maximum 3 at a time
2. Sequential Chunk Processing: Process one chunk at a time, completing all documents in a chunk before starting the next
3. Context Pause: After completing each chunk, provide a summary and pause for user acknowledgment
4. Progress Tracking: Display chunk progress (e.g., "Chunk 2/4: Processing TASKS-04, TASKS-05, TASKS-06...")

Why Chunking is Required:

• Prevents "Conversation too long" errors during batch processing
• Allows context compaction between chunks
• Enables recovery from failures without losing all progress
• Provides natural checkpoints for user review

Chunk Completion Template:

## Chunk N/M Complete

Generated:
- TASKS-XX: CODE-Ready Score 94%
- TASKS-YY: CODE-Ready Score 92%
- TASKS-ZZ: CODE-Ready Score 95%

Proceeding to next chunk...

Related Resources

• TASKS Skill: .claude/skills/doc-tasks/SKILL.md
• TASKS Validator: .claude/skills/doc-tasks-validator/SKILL.md
• Naming Standards: .claude/skills/doc-naming/SKILL.md
• Quality Advisor: .claude/skills/quality-advisor/SKILL.md
• TASKS Template: ai_dev_flow/11_TASKS/TASKS-TEMPLATE.md
• Implementation Contracts: ai_dev_flow/11_TASKS/IMPLEMENTATION_CONTRACTS_GUIDE.md

Review Document Standards (v2.2)

Review reports generated by this skill are formal project documents and MUST comply with shared standards.

Reference: See REVIEW_DOCUMENT_STANDARDS.md in the skills directory for complete requirements.

Key Requirements:

1. Storage Location: Same folder as the reviewed TASKS document
2. File Naming: TASKS-NN.R_review_report_vNNN.md (legacy) or TASKS-NN.A_audit_report_vNNN.md (preferred)
3. YAML Frontmatter: Required with artifact_type: TASKS-REVIEW, layer: 11
4. Score Field: code_ready_score_claimed / code_ready_score_validated
5. Parent Reference: Must link to parent TASKS document

Example Location (ALWAYS use nested folders):

docs/11_TASKS/TASKS-03_f3_observability/
├── TASKS-03_f3_observability.md         # ← Main document
├── TASKS-03.R_review_report_v001.md     # ← Review report
├── TASKS-03.F_fix_report_v001.md        # ← Fix report
└── .drift_cache.json                     # ← Drift cache

Nested Folder Rule: ALL TASKS use nested folders (TASKS-NN_{slug}/) regardless of size. This keeps companion files (review reports, fix reports, drift cache) organized with their parent document.

Version History