The Agent Skills Directory

[Prompt Injection] (LOW): The skill contains an indirect prompt injection surface (Category 8) where the agent is instructed to ingest and act on external data. Ingestion points: SKILL.md (Step 2) and DEPENDENCIES.md (Step 4) prompt the agent to process 'audit findings' and 'package.json' content. Boundary markers: Delimiters and warnings to ignore embedded instructions are absent from these interpolation points. Capability inventory: The agent has permissions to perform high-impact actions like npm uninstall, database schema modifications, and file system deletions (rm -rf node_modules). Sanitization: No sanitization or verification logic is provided for the ingested data.
[Command Execution] (SAFE): The skill utilizes common CLI tools such as grep, find, and curl for auditing purposes. These commands are directed at the user's own application environment and are consistent with the skill's optimization goals.
[External Downloads] (SAFE): The skill recommends several reputable third-party developer utilities (e.g., depcheck, artillery) from the standard npm registry. These are verified tools for the intended use cases.

optimize