Skills and rules
Every skill and rule, on one page
Looking for one by name? This is everything the toolkit ships, grouped the way the pages group them. Pick a name to jump to its page, or open its file on GitHub. The list is rebuilt from the toolkit repo every night.
Task workflow
Refine, plan, act, consolidate
- fetch-ticket downloads a ticket from Jira, GitHub, Azure DevOps or similar into a self-contained TICKET.md, with its attachments and a list of the tickets it links Read the SKILL.md →
- attach-to-ticket saves a screenshot or file you paste, or one you give by path or link, next to the ticket file and references it there, like an attachment the fetch could not download Read the SKILL.md →
- refine-ticket defines the WHAT: checks the ticket, or a raw idea, against the codebase, settles the open decisions with you and saves a REQUIREMENTS.md Read the SKILL.md →
- create-implementation-plan defines the HOW: settles the technical decisions with you and writes a PLAN.md a fresh session can execute Read the SKILL.md →
- harden-artifact checks a REQUIREMENTS.md or PLAN.md against the ticket, the code and the related tickets in a fresh session, and you decide each finding that survives Read the SKILL.md →
- split-plan-tasks breaks the plan into small tasks you can review and commit on their own, in groups that each make one PR, added to the end of PLAN.md as a checklist once you approve Read the SKILL.md →
- execute-plan-tasks runs one task of the PLAN.md checklist, then stops: it implements the task, runs your project's checks and ticks it off, so you review every task before the next one starts Read the SKILL.md →
- handover packages a finished change as a paste-ready PR description: what it does and why, the decisions worth knowing, where to look Read the SKILL.md →
- create-manual-test-instructions turns a ticket or its requirements into a MANUAL-TEST.md a non-author can follow Read the SKILL.md →
- review-ticket triages a ticket before anyone picks it up: a verdict, a feature walkthrough and the questions worth raising, saved in a TICKET-REVIEW.md Read the SKILL.md →
- verify-understanding a teach-back chat over a TICKET-REVIEW.md: you explain the feature in your own words, the agent probes and corrects Read the SKILL.md →
- check-ticket-implementation checks how much of a ticket is already built, marking each requirement done, partial, not done or not verifiable in a TICKET-STATUS.md Read the SKILL.md →
- grill-me interviews you about any plan, design or idea, one question at a time, each with a recommended answer, until every branch is settled Read the SKILL.md →
- prepare-prompt writes the prompt that hands this session's work to a fresh session, as short as the next session allows Read the SKILL.md →
PR reviews
Both sides of the review
- fetch-pr-review collects the comments reviewers left on your PR into a PR-REVIEW.md, ready to address or push back on Read the SKILL.md →
- refine-pr-review goes through a fetched review with you, comment by comment, drafting the replies and turning the accepted changes into a REQUIREMENTS.md Read the SKILL.md →
- review-code-assistant prepares your review of a PR or branch: concise comments with file and line, nothing posted Read the SKILL.md →
- maintainer-review reviews someone else's PR as the maintainer deciding whether it merges: every comment walked, every claim verified, nothing posted without your go-ahead Read the SKILL.md →
Fresh eyes review
Let a sub-agent review the code
- fresh-eyes-review a sub-agent with a clean context reviews a changeset and reports back what the session that wrote it misses Read the SKILL.md →
- self-review checks your branch before you ask anyone to review it: a sub-agent with a clean context reviews it the way a maintainer would, you go through each finding, and a short SELF-REVIEW.md goes into the PR Read the SKILL.md →
Context & memory
Keep the context lean
- context-checkup audits what auto-loads into a session before you type and suggests what to trim Read the SKILL.md →
- memory-doctor cleans up the memory your agent keeps accumulating on its own, moving what matters to the right place Read the SKILL.md →
- agentify-project gives a project a lean, agent-neutral setup: a short AGENTS.md index, shared skills and gitignore hygiene, re-runnable as an audit Read the SKILL.md →
Skill & doc authoring
Teach your agent
- compact-docs-writer writes or rewrites a doc to carry every rule and intent in the least text possible Read the SKILL.md →
- compact-skill-creator creates or edits skills, keeping them lean and effective Read the SKILL.md →
- self-improve captures a lesson into the skill, doc or check that should have prevented the mistake, so it doesn't repeat Read the SKILL.md →
Talking to humans
Texts that sound like you, explanations you understand
- use-conversational-language the voice for texts that should read as if a person typed them: PR comments, replies, commit messages, code comments Read the SKILL.md →
- explain-in-simple-language rewords answers, recaps and questions to you, and documents a teammate reads, so they are understood on the first read Read the SKILL.md →
Also in the toolkit
The SKILL.md says what each one does.
- extract-pr-review-rules Extract the conventions a repo's reviewers enforce in inline PR review comments into a per-reviewer rules doc; re-runs resume from a checkpoint.
- run-nx-checks Nx workspaces only formats, lints, tests and builds the projects your change affects, fixes what needs no judgment call and reports the rest. Not on Nx? Leave it out with
./install.sh --exclude run-nx-checks
Opinionated rules
Opt-in and installed separately: get them all with
./install-opinionated-rules.sh
or pick only the ones you want.
- compact-governing-docs every edit to a skill or governing doc goes through the compaction skills first Read the rule →
- git-read-only-by-default no commits, pushes or resets unless you asked for them
- no-ai-attribution your work stays yours: no AI co-author, no "generated with" footer
- no-nonsense-comments only comments a future reader with zero context still needs Read the rule →
- plans-directory where planning documents go: one folder per task in the project's planning directory, .agents/plans/ by default Read the rule →
- read-other-repos-governing-docs before the agent edits another repo, it reads and follows that repo's AGENTS.md or CLAUDE.md, since those load on their own only in the project you're working in Read the rule →
- self-contained-docs planning docs, and prompts that hand work to a fresh session, carry everything that session needs and nothing more Read the rule →
- self-improve-on-correction when you correct the agent on something a doc governs, it offers to capture the lesson with /self-improve Read the rule →
- write-realistic-texts texts other people read get the conversational voice, and you see the wording before anything is posted Read the rule →
- write-simple-explanations applies explain-in-simple-language whenever the agent asks you something, explains, recaps or writes a document a person reads Read the rule →
- use-my-mothertongue a copy-paste snippet, not a file: your agent talks to you in your own language, while everything it writes into the project stays in English