Use when repeated code review, doc review, design spec review, implementation plan review, or multi-agent review loops need severity gates, stop conditions, and implementation-readiness verdicts to prevent over-review, reviewer drift, duplicate findings, minor/nit churn, or infinite Claude/Codex/Gemini cycles.
Use when the user explicitly invokes $control-return or /control-return to re-apply the three-next-steps rule mid-session. A SessionStart hook auto-applies the rule at session start; the model must not invoke this skill on its own.
Use when the user wants to record, list, retire, promote or migrate lessons learned (/lesson); when a bug is fixed, an assumption is disproved or the user corrects the agent and work reaches a natural pause; and at wrap-up together with activity-logger.
Use when explaining one function or method in repository context, including its caller, contract, transformations, side effects, branches, tests, and modification constraints.
Use when understanding a repository or subsystem before reading individual files, including its purpose, runtime boundaries, entry points, representative flows, configuration, tests, and reading path.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: