The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform repository maintenance tasks, including listing existing skills and executing local validation scripts such as scripts/check-skill-integrity.py and scripts/check-surface-sync.sh to ensure repository consistency.
[EXTERNAL_DOWNLOADS]: The skill's documentation contains a reference to the mattpocock/skills repository on GitHub, identifying it as a design source for the skill's implementation patterns.
[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow that ingests external data to generate repository content, creating a potential surface for indirect prompt injection.
Ingestion points: The skill collects user-provided purpose statements, trigger phrases, and arguments during the interview phase. It also reads the templates/skill-template.md file and existing sibling skills from the .claude/skills/ directory.
Boundary markers: The process includes a dedicated interview phase (Phase 1) where the agent gathers all information and confirms a design brief with the user before performing any write operations.
Capability inventory: The skill uses the Write tool to create new SKILL.md files and the Bash tool to run integrity checks.
Sanitization: The skill enforces strict parity between the body and frontmatter using the check-skill-integrity.py script, ensuring that only declared flags and tools are permitted.

new-skill