The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/inspect_layout.py uses subprocess.run to execute Git commands (rev-parse, status, branch). These calls are implemented securely using argument lists without a shell, and are scoped to checking the status of local skill repositories.
[INDIRECT_PROMPT_INJECTION]: The skill acts on third-party skill files (other Agent Skills) during the audit and optimization workflow. While this constitutes an ingestion surface for potentially untrusted data, the risk is mitigated by the tool's use of rule-based regex linting and the requirement for explicit user-provided or environment-discovered paths.
Ingestion points: Files like SKILL.md, README.md, and skill.yaml from audited skill directories are read by scripts/lint_skill.py and scripts/inspect_layout.py.
Boundary markers: The workflow in SKILL.md defines distinct steps for classification and review, though it does not use explicit prompt delimiters for the ingested content.
Capability inventory: The skill can read/write files in skill directories (scripts/sync_installation.py, scripts/bump_version.py) and perform Git operations.
Sanitization: Content is analyzed using hardcoded regex patterns for structural and portability checks rather than being directly interpreted as instructions.
[REMOTE_CODE_EXECUTION]: The documentation references installation via npx skills add, which is the standard platform mechanism for the vendor lovstudio. The scripts themselves are written using only Python standard libraries to ensure portability and security.

lov-skill-optimizer