The Agent Skills Directory

[DATA_EXPOSURE]: The skill scans project files to generate documentation. It includes specific instructions in references/roles/researcher.md to explicitly avoid extracting sensitive information such as SQL column definitions, internal credentials, HTTP client configurations, and test code, focusing only on high-level behavioral facts.
[COMMAND_EXECUTION]: The skill uses the Bash tool for codebase discovery and to verify that research output files are generated correctly using the wc -c command. The usage is restricted to administrative and discovery tasks.
[PROMPT_INJECTION]: The skill processes untrusted content from the codebase being documented. This surface is mitigated by strict extraction rules for the researcher sub-agent and an adversarial review phase where a reviewer sub-agent identifies and corrects quality or safety issues in the generated draft before it is finalized.

technical-writer