The Agent Skills Directory

[SAFE]: The skill implements a comprehensive defense-in-depth verification methodology. It includes instructions for running language-specific test suites, build commands, and linting tools to ensure code quality and correctness.
[COMMAND_EXECUTION]: The skill uses subprocess.run() in scripts/validate.py to perform legitimate syntax and test checks (e.g., node -c, gofmt, pytest). These are standard development practices scoped to local file validation and do not involve untrusted remote sources.
[DATA_EXPOSURE]: The instructions explicitly include safety checks to prevent data exfiltration, specifically instructing the agent to grep for and remove sensitive strings like password, secret, api_key, and token before completion.
[OBFUSCATION]: No obfuscated strings, hidden URLs, or suspicious encoding techniques were found in any of the skill files.
[INDIRECT_PROMPT_INJECTION]: While the skill processes project files, it includes strict boundary instructions requiring the agent to show actual command output and evidence rather than summarizing or interpreting prose that could contain injections.

verification-before-completion