The Agent Skills Directory

Command Execution: The skill executes local scripts for environment validation and privacy checking, specifically setup_preflight and privacy-llm-check. These components are part of the framework's internal setup process.
Indirect Prompt Injection Consideration: The skill processes potentially untrusted data from GitHub issue bodies, comments, and external Gmail threads. It includes a proactive security measure instructing the agent to treat all external content as data rather than instructions, and it enforces a privacy check protocol before data is used.
Data Handling and Sanitization: To protect sensitive information, the skill implements a 'redact-after-fetch' and 'reveal-before-send' protocol via a dedicated privacy tool, ensuring that sensitive identifiers are managed carefully during the triage process.
Safety Guardrails: The skill incorporates multiple safety-first design patterns, such as using draft-only modes for email communication to ensure human oversight and implementing hard stops that prevent accidental invalidation of issues that already have allocated CVEs or published advisories.

issue-invalidate