AiatullaAiatulla/agent-kit/AGENTS.md

agent-kit

AGENTS.md2 starsChanged 46 days agoMITView on GitHub

This repo is a portable kit that configures AI coding agents (Claude Code, Gemini CLI) on any machine. It is not a project scaffold; scaffolding lives in fastapi-nextjs-starter. The general agent rules in global/AGENTS.md apply here too; this file only adds what is specific to this repo. Run scripts/verify.sh fast after every change and scripts/verify.sh full before every commit. Both must exit 0. - global/: loaded by every tool in every project. It must stay stack-free (no framework or library…

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 31,542 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 repository10 other files it gives its agents.
- global/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- commiterSkill
- dna-scannerSkill
- dry-kiss-rulesSkill
- full-dna-scannerSkill
- image-scannerSkill
- landing-architectSkill
- page-auditorSkill
- project-scannerSkill

The file

We split this AGENTS.md into the rules it gives an agent

8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific
could be automatic: a setting could enforce it

Very specific:
Run scripts/verify.sh fast after every change and scripts/verify.sh full before every commit.
line 9
Very specific:
Every rule exists in exactly one place.
line 19
Very specific:
Before adding a rule, grep global/ and stacks/ for it and edit the existing one instead.
line 20
Distinctive:
Do not edit vendored skill content.
line 30
Very specific:
No GNU-only flags; sed -i always takes a suffix argument so it works on BSD and GNU.
line 36
Very specific:
Test installs only with a temporary HOME.
line 42
Distinctive:
No absolute user paths anywhere in the repo.
line 43
Very specific:
Use git mv for moves so file history is kept.
line 48

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

Working on agent-kit

This repo is a portable kit that configures AI coding agents (Claude Code, Gemini CLI) on any machine. It is not a project scaffold; scaffolding lives in fastapi-nextjs-starter. The general agent rules in global/AGENTS.md apply here too; this file only adds what is specific to this repo.

Check before you finish

Run scripts/verify.sh fast after every change and scripts/verify.sh full before every commit. Both must exit 0.

Tiers

global/: loaded by every tool in every project. It must stay stack-free (no framework or library names) and global/AGENTS.md stays at most 150 lines.
stacks/<name>/: rules and checks for one stack, attached to a project directory by kit sync.
Project tier: lives in each project, never in this repo.

Every rule exists in exactly one place. Before adding a rule, grep global/ and stacks/ for it and edit the existing one instead.

Adding things

A tool: add one file adapters/<tool>.sh defining <tool>_detect, <tool>_install_global, <tool>_uninstall_global, <tool>_sync_project and <tool>_doctor. Nothing else may need to change.
A stack: add one folder stacks/<name>/ with rules.md (max 80 lines), lint/, verify.sh (fast and full modes) and registry.sh. Nothing else may need to change.
An own skill: add skills/<name>/SKILL.md; the folder name must equal the name: in its frontmatter.
A third-party skill: add it under vendor/skills/<name>/ and record it in vendor/skills-lock.json. Do not edit vendored skill content.

Shell constraints

Dependencies: bash, git, jq only; shellcheck is optional.
Bash 3.2 compatible: no declare -A, mapfile, readarray or ${var,,}.
No GNU-only flags; sed -i always takes a suffix argument so it works on BSD and GNU.
Shared helpers go in bin/lib.sh; adapters and scripts source it instead of duplicating logic.
Files the kit generates in other places carry a header: generated by agent-kit, version, do not edit.

Safety

Test installs only with a temporary HOME; never against your real ~/.claude or ~/.gemini.
No absolute user paths anywhere in the repo.

Workflow

Specs live in docs/specs/, plans in docs/plans/. Use git mv for moves so file history is kept. Bump VERSION (semver) when the installed or synced output changes.

Download now (ZIP)

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

What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free