The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines triggers for the agent to perform self-reflection and quality assessments. It does not include commands to bypass safety filters, disregard system instructions, or reveal system prompts.
[COMMAND_EXECUTION]: The provided scripts (check-code-quality.py, check-completeness.py, and check-security.py) are designed for static analysis. They use the ast module to inspect Python code structure and regex for other patterns. They do not execute the code they analyze or perform unsafe subprocess operations.
[EXTERNAL_DOWNLOADS]: No network activity, external downloads, or package installations are defined. The SKILL.md mentions that automation features are planned but not yet implemented, and the provided scripts are local files.
[DATA_EXFILTRATION]: The skill does not access sensitive files (like SSH keys or credentials) or environment variables. It processes input provided through command-line arguments or standard input.
[REMOTE_CODE_EXECUTION]: While the check-security.py script contains regex patterns to detect dangerous behaviors like curl | bash in other code, it does not implement such behaviors itself.
[SAFE]: The skill acts as a linter and quality assurance tool. Its operations are transparent and restricted to text analysis, presenting no significant security risk to the agent or the underlying system.

analyzing-response-quality