The Agent Skills Directory

[SAFE]: The skill is an analytical framework for documentation that operates on local files without requiring network access, sensitive credentials, or escalated privileges.\n- [COMMAND_EXECUTION]: The skill uses the Bash tool to run a local Python script (./scripts/spec_audit_lint.py) to check the structural integrity of specification documents. This script is part of the skill package and relies exclusively on the Python standard library.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied documents such as feature specifications and PRDs. Ingestion points: Specification artifacts located via Read, Glob, or direct user input. Boundary markers: The agent is directed to follow a specific adversarial rubric and output contract, though there are no explicit delimiters to prevent instruction injection within the audited text. Capability inventory: Uses Read, Grep, Glob, and Bash (for the internal lint script). Sanitization: None; input content is treated as the subject of logical analysis rather than instructions to be executed.

spec-auditor