Bootstrap sensible default agent skills into any repo and wire them across Claude, Cursor, Codex, and OpenCode.
Skill Claude CodeCodex
Review a PR, diff, or plan using reviewer and validator subagents in fresh context windows. Use for adversarial review, an independent second opinion, or a change-impact audit.
Skill Codex
Create or change repository-owned scheduled, webhook, or agent-run automations. Use for their playbooks, runtime boundaries, and delivery failures.
Skill Claude CodeCodex
Simplify existing code or a proposed change while preserving behavior. Use for an explicit cleanup or complexity review.
Skill Claude CodeCodex
Diagnose a bug, regression, flaky result, or unexplained slowness when its cause is not established.
Skill Claude CodeCodex
Explain how an unfamiliar subsystem works or identify where a change belongs. Use for code walkthroughs and ownership questions.
Skill Claude CodeCodex
Create, revise, or consolidate agent skills. Use when changing skill instructions, discovery descriptions, or supporting resources.
Skill Claude CodeCodex
Produce and compare independent solutions using fresh-context subagents. Use for an arena or a request for competing designs, implementations, or other substantial artifacts.
Skill Claude CodeCodex
Use when asked to create or update repo-local verification instructions for an actual app, CLI, or API, such as a verify-app skill. Do not use for ordinary implementation or one-off testing.
Skill Claude CodeCodex
Use when asked to prevent a recurring, evidenced repository mistake or repeated review correction. Do not trigger for a single speculative concern, ordinary bug fix, or personal memory mining.
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: