Define requirements for an OpenClaw integration
Document a proposed integration's data needs, actions, credentials, and review boundaries.
完了した決断に基づいてドキュメントを更新
---
name: cdd-doc-update-c68433518e8b
description: 完了した決断に基づいてドキュメントを更新
allowed-tools: Read, Edit, Write, Bash(cdd:*), Glob, Grep
---
# cdd-doc-update Command Template
You are updating project documentation based on completed CDD decisions.
## Your Task
Synchronize documentation with decisions that have `implementationStatus: DONE` but `docSyncStatus: NOT_SYNCED`.
## Process
### 1. Find Decisions Needing Documentation Sync
Use Bash tool to find all cdd.md files with:
- `implementationStatus: DONE`
- `docSyncStatus: NOT_SYNCED`
```bash
cdd status --detailed --implementation DONE
```
Then use Read tool to check each file's `docSyncStatus` field.
**Important**: Only process files that **have** the `docSyncStatus` field. Old files without this field should be ignored.
### 2. Read Each Decision and Extract Context
For each decision requiring sync:
1. Read the full cdd.md file
2. Understand what was implemented:
- **Goal**: What problem was solved
- **Selection**: What approach was taken
- **Context**: Important constraints and background
3. Extract the decision ID and title
4. **Extract gathered_context from Context section**
### 3. Identify Documents to Update
**Primary source: gathered_context in Context section**
If the cdd.md has gathered_context:
```yaml
gathered_context:
- path: docs/architecture/overview.md
summary: ...
relevance_to_task: |
CLIコマンド追加時は「ソースコード構造」セクションを参照。
実装完了後にdocs/architecture/overview.mdの更新が必要。
```
Use these paths as the guide for what to update. The `relevance_to_task` often indicates:
- Which sections need updating
- Why the document was referenced
- What changes might be needed
**If gathered_context is missing:**
Fall back to common documentation targets:
- `README.md`: User-facing features and usage
- `docs/architecture/overview.md`: System architecture and design decisions
- `docs/guidelines/development.md`: Developer workflows and practices
- `CHANGELOG.md`: Version history and notable changes
### 4. Update Referenced Documents
For each document in gathered_context (or fallback list):
1. **Read the document** to understand current content
2. **Identify the relevant section** based on relevance_to_task hints
3. **Add or update content** to reflect the implementation
4. **Preserve existing content** - never remove unrelated documentation
**Guidelines**:
1. **Find the right section**: Use relevance_to_task hints or Grep to locate where this information belongs
2. **Add new information**: Write about what was implemented and why
3. **Preserve existing content**: Never remove unrelated documentation
4. **Use clear formatting**: Headers, lists, code blocks as appropriate
5. **Link to cdd.md**: Reference the decision file for details
**Example Update**:
```markdown
## New Export Command (#EXPORT-001)
Added `cdd export` command for exporting decision data.
- Formats: JSON, CSV, Markdown
- Filters: by status, by date range
- Output: stdout or file
See [CDD/tasks/15-export-command.cdd.md](CDD/tasks/15-export-command.cdd.md) for decision rationale.
```
### 5. Update docSyncStatus
After successfully updating documentation:
1. Use Edit tool to change `docSyncStatus: NOT_SYNCED` to `docSyncStatus: SYNCED`
2. Only update this field - preserve all other frontmatter
### 6. Summary Report
After processing all decisions, report:
1. **Decisions processed**: List of IDs and titles
2. **Documentation updated**: Which files were modified (from gathered_context)
3. **Sync status**: Confirmation that docSyncStatus was updated
## Important Rules
- **Use gathered_context as guide**: These are the documents that were explicitly referenced
- **Never invent documentation**: Only document what was actually implemented
- **Verify implementation**: Check that code/features exist before documenting
- **Maintain accuracy**: Documentation must match current reality
- **Keep it concise**: Focus on high-level understanding, not implementation details
- **Link to cdd.md**: Let decision files serve as detailed source of truth
## File Editing Principles
When updating documentation:
1. **Read the entire documentation file first**
- Understand existing structure and sections
- Identify where new information belongs
2. **Preserve existing content**
- Add new sections or paragraphs
- Do not remove unrelated content
- Example: Architecture doc has sections A, B, C
- Adding section D → Result: A, B, C, D (NOT A, B, D)
3. **Use appropriate formatting**
- Match existing style and conventions
- Use headers to organize information
- Use lists for multiple items
- Use code blocks for examples
4. **Verify changes before finalizing**
- Ensure no content was accidentally removed
- Check that formatting is correct
- Confirm links and references work
## Integration with cdd-gather-context
The gathered_context in the Context section serves as your **document update guide**:
- **path**: The document to potentially update
- **summary**: What the document contains
- **relevance_to_task**: Why it was referenced and what might need updating
This creates a traceable flow:
1. `/cdd-gather-context` identifies relevant documents
2. Main agent adds them to cdd.md Context
3. `/cdd-review-decision` verifies the references
4. `/cdd-implement` does the implementation
5. `/cdd-review-implementation` reviews the code
6. `/cdd-doc-update` updates the referenced documents ← You are here
## Error Handling
**If no decisions need sync**:
- Report: "No decisions require documentation sync (all are SYNCED or lack docSyncStatus field)"
- Do not modify any files
**If gathered_context is missing**:
- Warn: "Decision lacks gathered_context in Context section. Using fallback document list."
- Proceed with common documentation targets
**If documentation file doesn't exist**:
- Ask user: "Documentation file X doesn't exist. Should I create it?"
- Wait for confirmation before creating
**If uncertain where to document**:
- Suggest options to user based on gathered_context hints
- Wait for guidance before proceeding
## After Completion
Present summary:
1. Number of decisions synced
2. Documentation files updated (mapped to gathered_context paths)
3. List of decision IDs marked as SYNCED
---
name: cdd-doc-update-c68433518e8b
description: 完了した決断に基づいてドキュメントを更新
allowed-tools: Read, Edit, Write, Bash(cdd:*), Glob, Grep
---
# cdd-doc-update Command Template
You are updating project documentation based on completed CDD decisions.
## Your Task
Synchronize documentation with decisions that have `implementationStatus: DONE` but `docSyncStatus: NOT_SYNCED`.
## Process
### 1. Find Decisions Needing Documentation Sync
Use Bash tool to find all cdd.md files with:
- `implementationStatus: DONE`
- `docSyncStatus: NOT_SYNCED`
```bash
cdd status --detailed --implementation DONE
```
Then use Read tool to check each file's `docSyncStatus` field.
**Important**: Only process files that **have** the `docSyncStatus` field. Old files without this field should be ignored.
### 2. Read Each Decision and Extract Context
For each decision requiring sync:
1. Read the full cdd.md file
2. Understand what was implemented:
- **Goal**: What problem was solved
- **Selection**: What approach was taken
- **Context**: Important constraints and background
3. Extract the decision ID and title
4. **Extract gathered_context from Context section**
### 3. Identify Documents to Update
**Primary source: gathered_context in Context section**
If the cdd.md has gathered_context:
```yaml
gathered_context:
- path: docs/architecture/overview.md
summary: ...
relevance_to_task: |
CLIコマンド追加時は「ソースコード構造」セクションを参照。
実装完了後にdocs/architecture/overview.mdの更新が必要。
```
Use these paths as the guide for what to update. The `relevance_to_task` often indicates:
- Which sections need updating
- Why the document was referenced
- What changes might be needed
**If gathered_context is missing:**
Fall back to common documentation targets:
- `README.md`: User-facing features and usage
- `docs/architecture/overview.md`: System architecture and design decisions
- `docs/guidelines/development.md`: Developer workflows and practices
- `CHANGELOG.md`: Version history and notable changes
### 4. Update Referenced Documents
For each document in gathered_context (or fallback list):
1. **Read the document** to understand current content
2. **Identify the relevant section** based on relevance_to_task hints
3. **Add or update content** to reflect the implementation
4. **Preserve existing content** - never remove unrelated documentation
**Guidelines**:
1. **Find the right section**: Use relevance_to_task hints or Grep to locate where this information belongs
2. **Add new information**: Write about what was implemented and why
3. **Preserve existing content**: Never remove unrelated documentation
4. **Use clear formatting**: Headers, lists, code blocks as appropriate
5. **Link to cdd.md**: Reference the decision file for details
**Example Update**:
```markdown
## New Export Command (#EXPORT-001)
Added `cdd export` command for exporting decision data.
- Formats: JSON, CSV, Markdown
- Filters: by status, by date range
- Output: stdout or file
See [CDD/tasks/15-export-command.cdd.md](CDD/tasks/15-export-command.cdd.md) for decision rationale.
```
### 5. Update docSyncStatus
After successfully updating documentation:
1. Use Edit tool to change `docSyncStatus: NOT_SYNCED` to `docSyncStatus: SYNCED`
2. Only update this field - preserve all other frontmatter
### 6. Summary Report
After processing all decisions, report:
1. **Decisions processed**: List of IDs and titles
2. **Documentation updated**: Which files were modified (from gathered_context)
3. **Sync status**: Confirmation that docSyncStatus was updated
## Important Rules
- **Use gathered_context as guide**: These are the documents that were explicitly referenced
- **Never invent documentation**: Only document what was actually implemented
- **Verify implementation**: Check that code/features exist before documenting
- **Maintain accuracy**: Documentation must match current reality
- **Keep it concise**: Focus on high-level understanding, not implementation details
- **Link to cdd.md**: Let decision files serve as detailed source of truth
## File Editing Principles
When updating documentation:
1. **Read the entire documentation file first**
- Understand existing structure and sections
- Identify where new information belongs
2. **Preserve existing content**
- Add new sections or paragraphs
- Do not remove unrelated content
- Example: Architecture doc has sections A, B, C
- Adding section D → Result: A, B, C, D (NOT A, B, D)
3. **Use appropriate formatting**
- Match existing style and conventions
- Use headers to organize information
- Use lists for multiple items
- Use code blocks for examples
4. **Verify changes before finalizing**
- Ensure no content was accidentally removed
- Check that formatting is correct
- Confirm links and references work
## Integration with cdd-gather-context
The gathered_context in the Context section serves as your **document update guide**:
- **path**: The document to potentially update
- **summary**: What the document contains
- **relevance_to_task**: Why it was referenced and what might need updating
This creates a traceable flow:
1. `/cdd-gather-context` identifies relevant documents
2. Main agent adds them to cdd.md Context
3. `/cdd-review-decision` verifies the references
4. `/cdd-implement` does the implementation
5. `/cdd-review-implementation` reviews the code
6. `/cdd-doc-update` updates the referenced documents ← You are here
## Error Handling
**If no decisions need sync**:
- Report: "No decisions require documentation sync (all are SYNCED or lack docSyncStatus field)"
- Do not modify any files
**If gathered_context is missing**:
- Warn: "Decision lacks gathered_context in Context section. Using fallback document list."
- Proceed with common documentation targets
**If documentation file doesn't exist**:
- Ask user: "Documentation file X doesn't exist. Should I create it?"
- Wait for confirmation before creating
**If uncertain where to document**:
- Suggest options to user based on gathered_context hints
- Wait for guidance before proceeding
## After Completion
Present summary:
1. Number of decisions synced
2. Documentation files updated (mapped to gathered_context paths)
3. List of decision IDs marked as SYNCED
Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.