aginit
AGENTS.md1 starsChanged 28 days agoMITView on GitHub
Set up new or existing projects for AI coding agents in one command: AGENTS.md, skills, codebase graph and test tooling.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository32 other files it gives its agents.
- .agents/skills/humanizer/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- ask-mattSkill
- codebase-designSkill
- code-reviewSkill
- diagnosing-bugsSkill
- domain-modelingSkill
- grillingSkill
- grill-meSkill
- grill-with-docsSkill
- handoffSkill
- humanizerSkill
- implementSkill
- implement-specSkill
- improve-codebase-architectureSkill
- karpathy-guidelinesSkill
- ponytailSkill
- prototypeSkill
- prSkill
- researchSkill
- retroSkill
- setup-matt-pocock-skillsSkill
- tddSkill
- teachSkill
- to-questionnaireSkill
- to-specSkill
- to-ticketsSkill
- triageSkill
- wait-whatSkill
- wayfinderSkill
- wizardSkill
- writing-for-agentsSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 7
Keep changes minimal, test-verified, and idempotent.
- Distinctive: line 8
Never commit broken builds, failed typechecks, or unverified changes.
- Very specific: line 10
Never add Node files (package.json, tsconfig.json) to a non-JS project root.
- Distinctive: line 22
When the user asks to "bootstrap this project" or kicks off a new initiative:
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
aginit
Set up new or existing projects for AI coding agents in one command: AGENTS.md, skills, codebase graph and test tooling.
Core Invariants
- Orchestrate upstream tools; do not duplicate, vendor, or fork them.
- Keep changes minimal, test-verified, and idempotent.
- Never commit broken builds, failed typechecks, or unverified changes.
- Tools and skills are registry entries in
src/registry/catalog.ts; never special-case one in CLI, writers or doctor. - Never add Node files (
package.json,tsconfig.json) to a non-JS project root.
Essential Commands
- Package Manager:
pnpm - Build:
pnpm build - Typecheck:
pnpm typecheck - Test:
pnpm test - Full check:
pnpm check(typecheck, tests, build, audit) - Packed-CLI smoke:
pnpm smoke(--frameworksalso scaffolds Vite and Next.js)
<!-- aginit:start — managed by npx @antomanc/aginit (re-run to refresh) -->Bootstrap & First Session
When the user asks to "bootstrap this project" or kicks off a new initiative:
- Understand product intent and ask 2-3 essential questions to resolve ambiguity.
- Draft or refine canonical context in
PRODUCT.mdandDESIGN.md(ordocs/adr/). - Plan in small verifiable increments and implement test-first using installed skills.
Guardrails (karpathy-guidelines)
- Think before coding: state assumptions and trade-offs; ask instead of guessing.
- Simplicity first: minimum code that solves the problem; no speculative features or abstractions.
- Surgical changes: touch only what the task needs; no silent scope creep or drive-by refactors.
- Goal-driven execution: define a verifiable success check first, then loop until it passes.
Task Routing
Load only what the task needs. Skills live in .agents/skills/ (linked into .claude/skills/ for Claude Code).
| Task | Read / Use |
| Automated unit & E2E testing | docs/agents/testing.md |
| Engineering workflow & skills | docs/agents/workflows.md |
| Scoping or simplifying a change | ponytail skill |
| Prose & Copy (docs, README, UI text) | humanizer skill |
Anti-Rationalization
| Excuse | Rule |
| "Unit tests pass, so the feature works." | Insufficient. Unit tests do not verify wiring or rendering. Exercise the feature end to end (integration or smoke run) before marking complete. |
<!-- aginit:end -->Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
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.