| name | review-repo |
| description | Perform a repo-wide code and documentation review focused on autonomy, agent swarm design, security, and decentralization; produce structured summaries and raw captures. |
SKILL: review-repo
Summary
Perform a repository-wide code and documentation review that focuses on alignment with the prime directive: building an autonomous agent swarm that self-improves, optimizes, evolves, and provides secure, decentralized networking to extend Copilot CLI and Copilot SDK.
When to run
- After
review-repo and before next-steps.
- Periodically during planning cycles or before major design decisions.
Inputs
- allow_run (bool, default: false) — when true, safe runtime checks and linters may be invoked.
- include_globs (array) — optional list of glob patterns to limit review scope.
- output_format (string):
text or json.
- safety_controls (optional) — configuration keys such as
director.allow_execute and corresponding env V_DAEMON_ALLOW_EXECUTE can be used to prevent runtime actions (default: disabled).
Outputs
- run/skills/review-repo//review.txt (readable stakeholder summary — informational only)
- run/skills/review-repo//review.json (structured results)
- run/skills/review-repo//raw/... (raw captures used to build the report)
Collection steps
- Capture git metadata and top-level layout.
- Preview core docs: README.md, AGENT.md, doc/, docs/.
- List scripts and capture shebangs/headers.
- Detect TODO/FIXME and capture context.
- Search for domain keywords: agent, swarm, autonomous, self-improve, evolve, optimize, decentralized, secure, copilot, supervisor, daemon, director.
- Produce a structured JSON with counts, file-type breakdown, and raw captures.