The Agent Skills Directory

[SAFE]: The skill provides developer tools and templates for deterministic logic verification.
No prompt injection or malicious instructions were found in the skill definitions or documentation.
The auditing scripts (check-determinism.mjs, check-golden-vectors.mjs) operate locally on the repository to verify code patterns and reproduction of test vectors.
Subprocess execution is restricted to internal self-tests and local git commands for reading committed state during audits.
Dynamic module loading is used exclusively for evaluating local generator functions against their expected test vectors during diagnostic runs.
The custom implementation of PNG processing in check-lib.mjs is used for managing test fixtures and does not contain malicious logic.

seeded-determinism-and-golden-vectors