The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local validation script at scripts/check.mjs. This script serves as a mechanical cross-check to ensure the generated requirements file correctly references all components identified in the codebase.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the codebase being analyzed. 1. Ingestion points: The entire codebase, including routes, public APIs, CLI commands, UI actions, and tests, is read via search and grep tools. 2. Boundary markers: The skill includes a 'Human approval' rule, specifying that the generated file remains a draft until explicitly reviewed and approved by a human. 3. Capability inventory: File system read access, search/grep operations, and execution of the local scripts/check.mjs script. 4. Sanitization: No specific input sanitization is mentioned; however, the structured output (YAML) and human-in-the-loop requirement mitigate the risk of direct command injection into the agent's logic.

requirements-discover