The Agent Skills Directory

[COMMAND_EXECUTION]: Executes standard development commands such as npm, pnpm, npx, pyright, ruff, git, and grep to verify project state. The instructions specifically use 'npx --no-install', which is a security best practice that prevents the installation of unknown packages at runtime.
[DATA_EXFILTRATION]: Includes commands to search for hardcoded secrets within the local repository using grep. This is a defensive security measure designed to help users identify and remove sensitive data before committing code, with no evidence of remote data transmission.
[PROMPT_INJECTION]: The skill is designed to ingest and report on the output of external tools like compilers and test runners. While this represents a surface for indirect prompt injection (where a malicious project could produce crafted error messages), the skill's logic is focused on status reporting and does not include unsafe processing of that data.

verification-loop