The Agent Skills Directory

Command Execution: The skill uses shell tools like rg (ripgrep) for searching and mktemp for managing temporary task notes. These operations are restricted to the primary documentation workflow and are handled within appropriate workspace boundaries.
Technical Verification Surface: The instructions suggest using project-specific validation tools, such as linters or tests, to verify code accuracy. The skill includes precautions against executing destructive commands and advises checking package scripts and imports beforehand.
Information Protection: The skill contains explicit directives to exclude credentials, secrets, or sensitive data from documentation examples and working notes, emphasizing the use of environment variables for authentication.
Indirect Prompt Injection Mitigations: The skill addresses the risk of malicious instructions within processed documents. (1) Ingestion points: Workspace documents and repository files (SKILL.md). (2) Boundary markers: Instructions explicitly state that source materials do not authorize unrelated actions. (3) Capability inventory: File searching, temporary file management, and project-specific validation. (4) Sanitization: Direct instructions to treat external content as data rather than authoritative commands.

technical-writer