The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automates various command-line tasks for repository maintenance:
Executes the repository-local script scripts/check-all.sh to validate the environment.
Uses pnpm exec vp to run build and check processes for the documentation site.
Employs git and the GitHub CLI (gh) for repository status, diffing, and branch management.
Step 10 specifies that the creation of the Pull Request, including committing and pushing, proceeds autonomously without additional user confirmation once the initial /finalize command is issued.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data that could potentially contain malicious instructions:
Ingestion points: Reads files from the skills/ directory to perform automated reviews and security checks.
Boundary markers: No specific delimiters or "ignore previous instructions" guards are implemented when the agent reads and processes the content of the skill files.
Capability inventory: The skill has significant capabilities, including modifying files (to fix findings), executing shell scripts, and performing network operations via git push.
Sanitization: The skill lacks an explicit sanitization layer to distinguish between legitimate data in skill files and potential instructions designed to influence the agent's logic during the review process.

finalize