The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes repository documentation content, which technically serves as an external data source. There is a potential consideration for indirect prompt injection if that content were to contain malicious instructions. However, the skill provides explicit mitigation in references/rai-guardrails.md, instructing the agent to treat all ingested content as data rather than instructions.
Standard Toolchain Execution: In references/validation-toolchain.md, the skill references standard repository maintenance commands (e.g., npm run lint:md, npm run docs:build). These are used within the context of the repository's existing CI/CD and development workflows for validation and build purposes.
Data Exposure Prevention: The skill includes a comprehensive sensitivity checklist in references/content-sensitivity.md to prevent the accidental inclusion of PII, secrets, or credentials in documentation. It also defines clear handoff triggers for formal security or privacy reviews when high-risk content is encountered.
Session State Management: The skill maintains session state locally within the .copilot-tracking/documentation/ directory, adhering to standard session persistence practices without externalizing data.

documentation