The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard development tools for verification, including git diff for version control review and bd create (Beads CLI) for issue tracking. These are legitimate operations within the scope of a developer-oriented skill.
[INDIRECT_PROMPT_INJECTION]: The skill functions by processing the agent's own output and task progress as data. While it includes checks for external data (like issue-classification in metadata), it lacks high-privilege tool interpolation or unsafe execution of untrusted data, presenting minimal risk.
[PROMPT_INJECTION]: The instructions use strong, imperative language such as "STOP. You are NOT done yet" and define Mermaid diagrams as "authoritative procedures" that must not be bypassed. This is an operational constraint designed to improve agent reliability and adherence to the verification protocol, rather than an attempt to override core AI safety guidelines.

verify-done