agentleFS
TsukiNi22TsukiNi22/ai-utils/skills/audit/audit-deps/SKILL.md

audit-deps

Skill1 starsChanged 3 days agoMITView on GitHub

Check the dependencies of a project (direct and transitive - CMake packages, pkg-config, FetchContent, submodules, vendored code, prebuilt binaries, includes, pip/npm/cargo) for their licenses - what must be credited, copyleft constraints, non-commercial / source-available / paid / proprietary terms, dependencies without license or of unknown origin - generate the third-party notices, and check their known vulnerabilities / compromised versions; result delivered as a PDF report. Use whenever the user asks what the dependencies allow, what to credit, if a dependency is legal/paid/restrictive, a THIRD_PARTY_NOTICES file, or a dependency security check.

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Not legal advice: say it once in the report.

    line 11
  2. Very specific:

    Check every unknown, proprietary, source-available, non-commercial, copyleft entry by hand: license file of the upstream repository (gh api …

    line 40

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “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

Dependencies: licenses, obligations, vulnerabilities

SKILL_DIR = directory of this file. Never remove or replace a dependency by yourself: report and propose. Not legal advice: say it once in the report.

1. Inventory + licenses

python3 SKILL_DIR/scripts/deps.py <root> --transitive --json /tmp/deps.json --md /tmp/deps.md
  • Finds the dependencies (CMake find_package / pkg_check_modules / FetchContent / ExternalProject, .gitmodules, vendored folders third_party/ vendor/ external/ libs/ cmake/libs/, prebuilt .a/.so, non standard #include <...>, requirements*.txt, package.json + lock, Cargo.toml + lock) and resolves each one to the system package owning it (rpm -qf / dpkg -S) for its version + license, or reads its LICENSE file.
  • --transitive: library packages required by the system ones (2 levels) and the lockfile entries; a transitive dependency inherits the scope of its parent (test when only used by the tests).
  • Category and obligation per dependency:
CategoryMeans for the project
public-domainnothing required
permissivecredit: copyright + license text with the distributed product
permissive-noticecredit + the NOTICE (Apache-2.0), state the changes
weak-copyleftchanges to the library stay open; LGPL: allow relinking (dynamic linking, or provide the objects when static)
strong-copyleftthe distributed product must be GPL-compatible and ship its sources
network-copyleftAGPL: sources also owed to network users (SaaS)
non-commercialno commercial use
source-availablenot open source (BUSL, SSPL, Elastic, Commons Clause...): production / SaaS / competition limits
proprietarypaid / EULA / redistribution limits: read the terms
unknownno license found = all rights reserved: not usable until clarified
system-runtimeglibc / libgcc / libstdc++: runtime exceptions, nothing to do

2. Review (the script is a first pass)

  • Check every unknown, proprietary, source-available, non-commercial, copyleft entry by hand: license file of the upstream repository (gh api repos/<o>/<r>/license), package metadata, headers of the sources.
  • Compound expressions (A AND B, A OR B): AND = all apply, OR = choose one (prefer the most permissive).
  • Linking matters: static (the user's libraries are .a) vs dynamic for LGPL; headers-only = the code is in the binary. Test-only dependencies are not distributed.
  • Compatibility with the project license (license skill, its compatibility table): ex. GPL dependency in an MIT product that is distributed = conflict.
  • "Illegal" / risky: no license, unclear origin (copied code, prebuilt binary without source/license), cracked or pirated components, assets (fonts, images, models) without rights, export-controlled crypto when relevant, paid SDKs used without a license (CUDA/NVIDIA EULA, Qt commercial, FMOD...).

3. Credits

python3 SKILL_DIR/scripts/deps.py <root> --notices THIRD_PARTY_NOTICES.md writes the attributions (name, version, license, home, license texts from /usr/share/licenses), to review and to ship with the product (and mention in the README, readme-style). Only on request: it creates a file in the project.

4. Vulnerabilities (also used by the audits)

python3 SKILL_DIR/scripts/vulns.py <root> --json /tmp/vulns.json --md /tmp/vulns.md [--recent-days 90]

OSV (PyPI / npm / crates.io / pinned git commits, malicious packages MAL-* included), pending Fedora security advisories of the installed packages (dnf updateinfo), GitHub security advisories of the upstream repositories compared with the installed version (gh). Advisories published in the last 90 days are flagged recent. For each finding: affected version, fixed version, the project usage (is the vulnerable feature used? grep it), action (update the package, bump the requirement, pin/replace). Debian/Ubuntu: also debsecan if installed.

5. Report

Through the report skill: ask the output format of report (Markdown + PDF recommended, PDF only, Markdown only), location audit/ at the root of the repository (or of the current folder outside a repository), never docs/ unless asked (audit/<YYYY-MM-DD>-dependencies.{md,pdf}), English unless asked. Sections: Summary (counts per category, blocking points first), Method (commit, tools, date, limits), Licenses (table per dependency: version, license, category, scope, obligation), Obligations (what to credit and where), Blocking / to clarify, Vulnerabilities (table by severity, recent first, fixed version, action), Sources. Then give the path(s).

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.