agent-skills

Create or maintain a skill library that groups two or more related, low-frequency Agent Skills behind one discoverable dispatcher. Use when the user wants to consolidate skills, reduce the active skill catalog, group manual runbooks, or keep exceptional workflows available without loading every skill description. Also use when deciding which skills should remain standalone versus move into a library. Triggers: skill library, skill of skills, meta-skill, consolidate skills, group skills, runbooks library, reduce active skills, skill catalog pollution, nested skills, skill dispatcher.

Evaluate a subagent by running its scenario suite as parallel subagents, scoring results against observable criteria, and producing exact FIND/REPLACE edits for any failing behaviors. Use this skill whenever the user wants to evaluate, test, score, or quality-check a subagent — even if they don't use those exact words. Triggers on: "evaluate the agent", "run evals on [agent]", "assess agent performance", "score the agent", "check if the agent works", "run the eval suite", "test the agent", "quality-check [agent]", "is the agent working correctly", or after applying fixes "did that improve things?". Also triggers proactively after an agent file is edited and the user asks "does it work now?" or similar. Triggers: evaluate agent, run evals, assess agent, score agent, test agent, quality-check, eval suite, agent performance, agent working, improve agent, rerun evals, benchmark agent.

Create, update, review, and execute manual Markdown implementation plans when harness planning mode is not being used. Use when the user asks for a plan file, manual plan, implementation plan, execution plan, roadmap, task checklist, planning document, or agent-maintained plan with statuses, validations, milestones, approval gates, and cleanup steps.

Port a repo-specific skill into a generic, reusable skill for the agent-skills library. Use this skill whenever the user asks to extract a skill from a project, generalize a skill, port a skill to another repo, convert a project-specific skill into a reusable one, or remove repo-specific context from a skill. Also use when the user mentions skill extraction, skill generalization, or creating a library skill from a project skill — even if they don't explicitly say "extract" or "generalize." If a skill exists in a project's `.agents/skills/` directory and the user wants to make it reusable across repositories, this skill should be used. Triggers: extract skill, generalize skill, port skill, convert skill, remove repo-specific, make skill reusable, skill extraction, skill generalization, library skill, repo-agnostic skill, skill from project.

Captures lessons from completed tasks and turns them into concrete improvements. Triggers proactively after high-ceremony tasks (build failures, multi-hour sessions, production incidents) and on-demand when the user asks for "post-mortem", "reflect", "improve the skill", "lessons learned", "what went wrong". Covers skills, repo artifacts (scripts, CI, configs), and process changes. For small/obvious fixes, edits directly. For structural changes, produces a plan. Use this whenever a task surfaced gaps in tooling, documentation, or automation.

End-to-end management of TODO items in a Markdown-based backlog. Covers creation with auto-assigned IDs, status tracking, archival of completed items, and format validation. Use this skill whenever the user asks to add a new TODO, check TODO status, archive completed items, validate the TODO system, or set up a new TODO backlog from scratch. Also use when the user mentions a task list, backlog, roadmap, or task tracking in a Markdown file — even if they don't explicitly say "TODO". If the repo has a `TODO.md` file or similar Markdown-based task tracking, this skill should be used to interact with it. Triggers: TODO, todo item, add todo, create todo, todo status, archive todo, validate TODO, TODO ID, TODO-0001, check TODO, todos, list todos, todo list, backlog, task tracking, roadmap, task list.