headless-claude — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited headless-claude (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.
The flags that make claude -p actually behave in scripts, agents, and cron.python3 headless_claude.py "summarize CHANGES.md in 3 bullets" # prints the command (dry)
python3 headless_claude.py --apply "summarize CHANGES.md in 3 bullets" # runs it
python3 headless_claude.py --apply --cwd ~/proj "fix the failing test" # in a repoA bare claude -p in an automated context can hang on a permission prompt, fire your editor/notify/Stop hooks, and load your whole user memory (slow). For unattended use you almost always want:
| flag | effect |
|---|---|
--setting-sources project | don't load USER ~/.claude hooks/settings; skip auto-memory → fast, no side effects |
--permission-mode bypassPermissions | no interactive approval gate — a script can't answer one |
neutral --cwd | don't drag in a project's CLAUDE.md / file tree |
Escape hatches: --allow-hooks (keep your hooks), --safe (let it prompt). BYOK — uses your logged-in plan, no API key.
claude / nonzero exit / timeout → FATAL + exit 2.claude CLI on PATH, logged in. No Python packages.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.