repo-init — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited repo-init (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Lay the foundation for a fresh, nearly empty, or never-organized repo. Scaffold the structure and leave placeholders — do not write product, plan, or design content (that's the user's and /rad-planner:plan's job). Use the templates in <plugin-root>/templates/ (where <plugin-root> is either ${CLAUDE_PLUGIN_ROOT}, global config plugins/rad-repo-manager, or local workspace plugins/rad-repo-manager).
Use this only for greenfield setup. An established repo with drift goes to /rad-repo-manager:repo-align instead.
git status --short
git remote get-url origin # for the AGENTS.md remote line, if setGlob the repo. Detect the app type and package manager if obvious (package.json, lockfiles, pyproject.toml, Cargo.toml, framework config) — this only informs the AGENTS.md baseline, not product content.
Confirm this is greenfield: no AGENTS.md and no real docs/prd.md/docs/plan.md. If the repo is already established, stop and recommend repo-align.
If --agents was passed, use it. Otherwise ask (via AskUserQuestion or ask_question on Antigravity): which coding agents will use this repo — Claude, Codex, Gemini, or a mix? This decides which shims to create. Codex reads AGENTS.md natively (no shim). If unclear, ask before creating non-Codex shims.
Create only files that don't already exist. Never overwrite a user-authored file without explicit confirmation.
Core active docs:
AGENTS.md ← templates/AGENTS.md (operating manual; placeholders for the user)docs/prd.md ← templates/prd.md (skeleton)docs/plan.md ← templates/plan.md (stub pointing at /rad-planner:plan)docs/handoff.md ← templates/handoff.md (snapshot shape, placeholder values)Agent shims (per scope; each a thin @AGENTS.md import — never duplicated content):
CLAUDE.md ← templates/CLAUDE.mdGEMINI.md ← templates/GEMINI.mdMinimal folders:
docs/reference/README.md ← templates/reference-README.md (the closed catalogmenu; do not create the reference docs themselves yet)
docs/archive/README.md ← templates/archive-README.md (history banner)Do not create docs/status.md, docs/roadmap.md, docs/implementation-plan.md, docs/inbox/, a top-level docs/design.md (it appears only when design work starts), loose root-level handoff/status/audit docs, or folder-specific AGENTS.md/CLAUDE.md/GEMINI.md.
Set the remote URL line in AGENTS.md from git remote get-url origin if available. Stamp each scaffolded doc's **Updated:** <YYYY-MM-DD> with today's date (the freshness scans key off it). Leave every other <PLACEHOLDER> section for the user. Do not invent goals, roadmap, architecture, decisions, or validation promises.
Tell the user, plainly: the structure is in place, the active read path is the four core docs, and the next steps are (a) run /rad-planner:plan to create the plan and (b) fill the <PLACEHOLDER> sections in AGENTS.md and docs/prd.md.
Repo init:
Detected repo type:
Agent scope:
Files created:
Files skipped (already exist):
Active read path:
Next step for the user:
Notes:<plugin-root>/templates/ — the doc-model skeletons<plugin-root>/references/doc-model.md — the tiers, the active core, the reference catalog~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.