Browse Skills

Loop Constraints reads loop-constraints.md at the start of every run and enforces all rules before triage or actions. Use it to apply binding guardrails such as denylist paths, push/merge checks, and test requirements.

Loop Triage reviews recent CI failures, commits, issues, and conversations, then produces a concise, prioritized findings report for an AI loop to act on. Use it to separate high-priority work from noise and persist state between runs.

Loop Triage reviews recent CI failures, commits, issues, and conversations, then produces a concise, prioritized findings report for an agent loop to act on. Use it to keep a coding agent's state file up to date with high-priority items, watch items, and noise.

Loop Verifier provides independent verification of loop-produced changes by running tests and checking diff scope. Use it after an implementer sub-agent to reject changes unless evidence is strong.

Loop Verifier reviews release note drafts against the changelog scan data to flag hallucinations, missing breaking or security items, and tone problems. Use it after draft-release-notes to independently check accuracy before human review.

Minimal Fix produces the smallest possible code change that fixes a specific, well-scoped issue such as a CI failure, reviewer comment, or typo. Use it when the fix target is explicit and avoid refactoring unrelated code.

Post-Merge Scan reviews recent merges to main for tech debt, TODOs, debug code, and small cleanup opportunities, then produces a prioritized fix list. Use it after merges to keep the codebase clean without manual inspection.

PR Review Triage tracks open pull requests, CI status, review staleness, merge conflicts, and unanswered review comments to produce a prioritized watchlist. Use it to surface blockers and decide the next action for each PR.

Fireworks Tech Graph creates geometry-checked technical diagrams from natural language, including architecture, flow, sequence, UML, and C4 visuals. Use it to export SVG, PNG, GIF motion, or offline interactive HTML for system and engineering concepts.

Actions Failure Analyzer investigates a failed GitHub Actions workflow run and identifies the specific root cause. It checks the job's recent history for flakiness, traces the breaking commit, and searches for open PRs that may already fix the issue.

Half Clone clones the later half of the current Claude Code conversation, discarding earlier context to reduce token usage while preserving recent work. Use it when a session has grown long and you want to continue with only the latest context.

Handoff writes or updates a HANDOFF.md document so the next AI agent with fresh context can continue the work. Use it to record goals, progress, what worked, what didn't, and next steps before handing off a coding session.