skills
AGENTS.md9 starsChanged 19 days agoMITView on GitHub
This repo is the skills. Every change is prose in a SKILL.md or a supporting reference file. The host invokes the installed plugin, not this checkout. A confirmed /grill-with-docs Read-back that names a change is edited in this checkout: glossary, ADR, and the skill edit in one commit, which also closes the Issue the session started from as done. Its Changelog entry follows in a changelog: commit of its own: under # Changelog in CHANGELOG.md, newest first, a ## yyyy-mm-dd:…
Files
mvdmio Skills
This repo is the skills. Every change is prose in a SKILL.md or a supporting reference file. The host invokes the installed plugin, not this checkout.
A confirmed /grill-with-docs Read-back that names a change is edited in this checkout: glossary, ADR, and the skill edit in one commit, which also closes the Issue the session started from as done. Its Changelog entry follows in a changelog: commit of its own: under # Changelog in CHANGELOG.md, newest first, a ## yyyy-mm-dd: <title> heading of at most six words, then one paragraph of at most three sentences a user of the skills reads once — what changed for them, and what went wrong before. A change only an agent or a maintainer of this repo notices — a Grilling profile, a /retro edit — gets no entry. A /triage problem whose ending would be Spec is edited the same way. Outside that exception, /to-spec and /implement stay unused. Read and follow skills/retro/SKILL.md, then point the installed plugin at the commit those steps left: push master, then run claude plugin marketplace update mvdmio and claude plugin update skills@mvdmio (on Grok, grok plugin update). Both hosts install from origin, so an unpushed commit never reaches them. A Read-back that names no change writes no Spec, keeps glossary entries already written, and still runs /retro.
A session the grill-queue runs unattended from a needs-grilling Issue is the exception: it edits no skill. Glossary and ADR changes still commit and push to master, then /to-spec rewrites the Issue as a Spec labelled ready-for-agent, and Implementer makes the skill edit with /implement.
Editing agent documents
Match skills/writing-for-agents/SKILL.md on every skill file, AGENTS.md, or CLAUDE.md you edit. Run /writing-for-agents on them after.
Human-facing skill output runs at skills/technical-writing/SKILL.md — see ADR-0009.
CONTEXT.md is the glossary for how the skills talk about the documents they read and write, and about each other. Use its terms exactly; add to it when a change coins one. Rules belong in an ADR.
In this repository26 other files it gives its agents.
- verifySkill
- brainstormSkill
- code-reviewSkill
- doctorSkill
- domain-modelingSkill
- explainSkill
- grillingSkill
- grill-with-docsSkill
- hillclimbSkill
- implementSkill
- learn-habitsSkill
- prototypeSkill
- prSkill
- researchSkill
- resolving-merge-conflictsSkill
- retroSkill
- setupSkill
- simulate-first-usersSkill
- technical-writingSkill
- to-specSkill
- triageSkill
- validate-specSkill
- wayfinderSkill
- what-did-i-get-doneSkill
- whySkill
- writing-for-agentsSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.