The Agent Skills Directory

[SAFE]: The skill provides a structured methodology for agent planning without introducing security vulnerabilities. The instructions focus on refining intent and managing technical debt through established engineering practices like BDD and Design-by-Contract.
[COMMAND_EXECUTION]: The provided scripts/validate.sh performs static analysis using grep to verify the integrity and versioning of the skill's instructions. These checks are local and do not perform network operations, privilege escalation, or access sensitive user data.
[INDIRECT_PROMPT_INJECTION]: The skill describes a surface for ingesting external data (caller trackers, existing intent, and source code). It mitigates risks associated with this ingestion by prescribing strict behavioral formatting (Given/When/Then), established domain vocabulary, and requirement for human-in-the-loop authority for scope changes.

plan