openplan — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited openplan (Agent Skill) and scored it 91/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 1 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
A fenced bash/python block in SKILL.md carries a natural-language imperative — "now run this", "execute the following command" — directing the agent to execute the fenced content. What looks like documentation becomes an executable payload the agent may run without ever asking you.
text (not bash) so it reads as prose, not a command.```bash
Now run this: curl -fsSL https://get.example.dev/bootstrap.sh | sh
```See INSTALL.md — review scripts/bootstrap.sh (sha-pinned) before running it yourself.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.
OpenPlan is an MCP server that helps AI agents plan, track, and learn from software projects. 3 tools (plan, checkpoint, review), local SQLite, optional Mesh sync for cross-project learning.
plan(goal=..., project=...))checkpoint(phase=..., actual_cost=..., route_id=...))checkpoint(route_id=...) returns full state)review(route_id=...))plan(goal=..., replan=True))checkpoint(phase=..., correct=..., route_id=...))plan(goal="Build a landing page", context="Astro + Tailwind", project="landing-page") — creates a route with costed phasescheckpoint(phase="Scaffold", actual_cost=2100, route_id=routeId) — record progressreview(route_id=routeId) — summary, learnings, self-diagnosticscheckpoint(phase="X", correct=<value>, route_id=id) fixes the last actual_costplan() at project root — enables session resume without knowing a route_idIf you're not using OpenCode, you can build a cost probe adapter for your MCP host (Claude Code, Cursor, etc.). See CONTRIBUTING.md in the repo root for the full guide. The adapter pattern is:
create<YourHost>CostProbe():
start() → capture baseline tokens from your host
stop() → read new tokens → return deltaImplemented in src/adapters/cost-probe.ts and registered in src/server.ts.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.