DeepChat SDD Documentation Cleanup Skill
Safely clean, prune, and tidy DeepChat SDD documentation after implementation and validation.
The skill explicitly requires explicit developer request to run and has a 'Never Delete' list (active work, unresolved clarification, referenced docs), showing care against data loss. However, it authorizes direct file deletion without explicit user confirmation or rollback mechanism. gh check is only used when available, but no clear permission scope limitation. Publisher unverified. Deductions: lack of explicit confirmation, rollback, and clearer permission boundaries.
Provides clear steps and validation commands, but static review without executing key paths. Depends on external tools (find, rg, gh) without handling unavailability except gh, external dependency availability not guaranteed. Error handling limited, no defined behavior on abnormal input (e.g., nonexistent directories). Deductions: no executed tests, thin error handling, insufficient external dependency guarantees.
Trigger conditions are precisely defined (only when explicitly asked) and non-fit scenarios (not automatic step) clear. However, capability boundaries and non-fit ranges are not exhaustively declared (e.g., which doc types are suitable or not), no examples. Environment fit is good, no reliance on overseas services, suitable for Chinese users. Deductions: incomplete boundaries and non-fit ranges, lack of examples.
Structured workflow and rules provided, including dependencies (gh) and validation steps. Missing installation notes (not standalone), FAQ, changelog, versioning, and maintenance responsibility. Publisher unverified, maintenance unclear. Deductions: no versioning, changelog, maintenance responsibility, FAQ.
Goals and benefits clear, cleanup rules detailed. But static review, outputs not verified, and evidence for comparative benefit (over manual) limited. Deductions: unverified outputs, insufficient comparative benefit evidence.
Validation steps and cleanup rules provided, but no test suite or reproducible third-party execution evidence. CI workflows exist (packaging), but do not cover this skill's key path. Deductions: no tests or reproducible execution evidence.
- Skill deletes files directly; ensure explicit developer request and check 'Never Delete' list before running.
- When gh is unavailable or unauthenticated, may incorrectly keep or delete docs; rely on local code and test evidence instead.
- No rollback mechanism provided; deletion is irreversible, recommend backup before use.
What does this skill do, and when should you use it?
This skill cleans up stale SDD (Spec-Driven Development) documentation in the DeepChat project. It scans docs/features, docs/issues, and docs/architecture directories, identifies completed or obsolete documents, and removes them. It prefers multi-agent review and verifies completion status by checking linked GitHub issues. The skill runs only when explicitly requested by the user, not as an automatic step in regular development workflows.
Scans docs/features, docs/issues, and docs/architecture directories. Inspects docs/spec-driven-dev.md, docs/README.md, and git status. Orchestrates parallel sub-agent reviews of documentation directories. Applies changes only with clear evidence and maintains a keep/delete/update list. Deletes plan.md and tasks.md files for completed feature or architecture goals. Deletes completed issue folders tied to closed GitHub issues. Deletes folders for removed features and obsolete architecture. Updates historical feature specs affected by architecture refactors. Runs rg and git status for validation.
- A developer who has just implemented a feature or architecture goal wants to clean up outdated plan files.
- A developer who has fixed a bug wants to delete the related issue documentation now that the GitHub issue is closed.
- A developer who has refactored an architecture wants to tidy the docs directory while preserving specs that are still valid.
- A developer who has removed a product feature wants to remove the corresponding documentation folders.
What are this skill's strengths and limitations?
- Helps keep docs tidy and up-to-date by removing stale files.
- Multi-agent review via parallel passes enhances efficiency.
- Includes safety guards against deleting important docs, such as active work and docs with unresolved clarification markers.
- Verifies GitHub issue status via GitHub CLI when available.
- Requires explicit user request; does not run automatically, which may limit its usefulness.
- Relies on GitHub CLI for issue status verification; without it, verification is impossible.
- Specific to DeepChat project documentation structure, limiting general applicability.
- No test suite; must be reviewed carefully to avoid accidental deletions.
How do you install this skill?
Place this skill folder (containing the SKILL.md) into your agent's skills directory. For example, for Claude Code, place it in ~/.claude/skills/.
How do you use this skill?
Explicitly request SDD cleanup in your prompt, e.g., "Please clean up stale SDD docs" or "Tidy the docs directory." The skill will not run automatically.