agentleFS
NAVANEETHVVINODNAVANEETHVVINOD/agent-orchestrator/skills/plugin-safety-review/SKILL.md

plugin-safety-review

Skill0 starsChanged 42 days agoApache-2.0View on GitHub

Audit a named plugin, skill, MCP server, or installer for provenance, permissions, instruction scope, dependencies, and concrete security risks before adoption or when the user requests a safety review.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Start with narrow read-only inventory.

    line 12
  2. Distinctive:

    Treat installation presence separately from connection and successful runtime use.

    line 16
  3. Distinctive:

    Use official source, release metadata, exact npm/PyPI version, immutable commit, and applicable public advisories.

    line 18
  4. Distinctive:

    Don't remove plugins, change connector permissions, or rewrite third-party cached skills during an audit unless the user also requests …

    line 33

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Plugin safety review

Resolve the named integration and exact installed version first. A skill is model guidance; scripts and MCP servers can execute code; remote connectors send data to another service. Review each boundary that actually applies to the request.

Start with narrow read-only inventory. Parse configuration and print only required non-secret fields; never dump environment variables, credential files, session logs, auth stores, private browser profiles, or unredacted process command lines. Inspect requested source/manifests without running installers, hooks, or package code. Treat installation presence separately from connection and successful runtime use.

Use official source, release metadata, exact npm/PyPI version, immutable commit, and applicable public advisories. If downloading source, keep it in a scratch folder. Record publisher, version, commit and integrity metadata. Do not claim provenance or a zero-advisory result proves the package is safe. Do not execute a package merely to identify it. A lockfile audit covers dependencies, not cloud/backend behavior.

Trace command flags and installer mutations, executable hooks, privilege and filesystem scope, shell escape paths, data sent off-machine, credential storage, remote transport auth, update policy, and collisions with existing skill folders. In instruction files, look for always-on routing, policy/authority overrides, automatic installs or publishing, secret reads, broad cleanup, and output that hides failures. Explain concrete impact.

Findings must include severity, affected version/configuration, source or local line evidence, trigger/preconditions, and a proportionate fix. Separate confirmed bugs, expected but risky capabilities, suspicious candidates, and unverified areas. Don't remove plugins, change connector permissions, or rewrite third-party cached skills during an audit unless the user also requests those changes. Prepare a concrete remediation for review when a change requires additional authorization.

Recommend the smallest needed tool set. Prefer native tools where sufficient, reviewed fixed versions and scoped installation where an external tool is needed. Preserve unrelated edits. Describe scan scope and limitations; do not certify safety.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.