Agents md init

Skill K95M65/AI_ONBOARD/skills/agents-md-init

Portable agent workflow framework for Codex, Claude Code, and OpenCode—shared project context, focused skills, independent review, and safe updates.

Install

npx -y skills add K95M65/AI_ONBOARD --skill agents-md-init

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

19 days oldThe repository was created 19 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Bootstraps an AGENTS.md file for a project by detecting its stack, package manager, and build/test/lint commands. Use when a repo has no AGENTS.md (or CLAUDE.md) yet and the user wants to onboard AI tools, or asks to create/generate agent instructions for the project.

SKILL.md

2.7 KB, as published. Nobody here has run it

AGENTS.md init

Create a first-draft AGENTS.md grounded in what the repository actually contains, not in guesses.

When to use

A project has no AGENTS.md yet and the user wants to set up AI-tool onboarding, or asks you to "generate an AGENTS.md / agent instructions" for the repo.

Steps

Detect the stack and commands — run the detector instead of assuming:
```
bash scripts/detect-stack.sh
```
It reports the language/ecosystem, package manager, and the real build / test / lint scripts it finds in manifests (package.json, pyproject.toml, Makefile, go.mod, Cargo.toml, etc.). For Apple projects it also reports Xcode projects/workspaces, SwiftPM, Tuist, Swift version files, shared schemes, and formatter/linter configuration.
Skim the structure — list top-level dirs and note what each holds (src/, tests/, packages/…). Read the existing README.md for setup steps and conventions already documented.
Draft the project-root AGENTS.md with these sections:
- Overview (stack, what it is) — 1–2 sentences.
- Setup, build, test, lint — use the exact commands the detector found.
- Structure — the real directories.
- Conventions — only ones you can verify (formatter config present, lint rules, naming actually used).
- For Swift/Apple projects — the project system, shared schemes, supported platforms and deployment targets, Swift language/concurrency settings, exact destinations used by CI, generated-project ownership, and distribution/signing references. Never copy signing identities or credential values.
Only claim what's true. If you can't confirm a convention from config or code, leave it out rather than inventing it. A wrong AGENTS.md misleads every agent that reads it.
Keep it short. It's loaded every session. Cut anything the code already makes obvious.
Offer to wire up tools — mention that CLAUDE.md can @AGENTS.md, GEMINI.md can symlink to it, and Codex/opencode/Cursor read it natively. (See the tool matrix / link.sh in AI_ONBOARD.)

Rules

Prefer editing an existing AGENTS.md over clobbering it — merge, don't overwrite.
Don't include secrets, tokens, or environment-specific paths.
Commands must be copy-paste runnable; verify names against the manifest, don't paraphrase.