← ClaudeAtlas

project-onboardinglisted

Use when initializing, auditing, or updating a project's AGENTS.md, onboarding Codex to a repository, capturing repo-specific commands, or turning sparse project context into high-signal agent instructions.
yuri-semenenko/ai-engineering-workspace · ★ 1 · AI & Automation · score 70
Install: claude install-skill yuri-semenenko/ai-engineering-workspace
# Project Onboarding Create or improve a project-level `AGENTS.md` that helps Codex work in the repo without restating generic engineering advice. ## What To Capture - Scripts: build, test, lint, typecheck, dev, format, and workspace flags. - Single-test command with a concrete example. - Migrations: location, generate/apply/rollback commands, Supabase workflow if relevant. - Local DB and seed data setup. - Required env vars and where values come from. - Non-obvious repo structure, server/client boundaries, route conventions, and generated files. - Auth, RLS, test harness, PR/commit conventions, CI gotchas, and monitoring only when they exist. ## What To Skip - Full file trees. - Generic "write tests" or "use TypeScript" advice. - Stack summaries already visible in `package.json`. - Friendly boilerplate. ## Process 1. Read existing `AGENTS.md`, `README`, `package.json`, test config, migration folders, and CI files. 2. Identify commands and gotchas from repo evidence, not guesses. 3. Keep `AGENTS.md` around 80-150 lines unless the repo genuinely needs more. 4. Prefer repo-specific facts over universal preferences. 5. If the repo already has good instructions, patch only stale or missing sections. Use `$CODEX_HOME/references/memory-seed/project_claude_md_checklist.md` for the longer migrated checklist when needed.