The Agent Skills Directory

[SAFE]: The skill operates primarily as a verification and reporting tool. It reads existing documentation (READMEs, test cases) and maps them to automated tests found in the repository.
[REMOTE_CODE_EXECUTION]: No direct execution of remote code was detected. The skill specifically states it does not run tests itself but delegates execution to a local 'quality-check' mechanism, which is a standard internal workflow for this environment.
[DATA_EXFILTRATION]: No network operations or sensitive file access patterns were found. The skill only writes a markdown report (criteria-coverage.md) back to the project repository.
[PROMPT_INJECTION]: The instructions are clear and follow a strict operational logic. There are no attempts to bypass safety filters or override system prompts.
[OBFUSCATION]: The content is provided in plain text and standard Markdown. No base64, zero-width characters, or homoglyph attacks were identified.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data (user-defined acceptance criteria and test cases). However, the output is restricted to a structured Markdown report, and the skill includes boundary markers/fingerprinting for idempotency, which limits the impact of potentially malicious input.

coverage-verify