The Agent Skills Directory

[Command Execution]: The skill uses the GitHub CLI (gh) and internal Python scripts to manage issue creation and project board status. It implements rigorous defensive patterns, such as using a dedicated 'Write tool' to store untrusted strings (like email subjects) in temporary files before passing them to the CLI via file-reading flags (e.g., gh api -F). This effectively prevents shell command injection attacks.
[Indirect Prompt Injection Surface]: As an intake tool, the skill processes untrusted email content that could contain malicious instructions. The skill addresses this by instructing the agent to treat all external content as data only, never as instructions. It also includes a sanitization step for markdown to prevent rendering exploits, such as demoting headings and defusing tracking pixels.
[Data Handling]: The skill facilitates the movement of sensitive security report data. It enforces a 'Golden Rule' of confidentiality, ensuring that data from private mailing lists is only moved to a private GitHub repository and never leaked to public surfaces. It also incorporates a privacy-focused redaction protocol (PII redaction) before processing report bodies.
[Environment Integrity]: The skill uses a pre-flight checker (setup_preflight) and hash verification (surface_hash) to ensure the execution environment and skill instructions remain consistent and authorized. This is part of the established framework for managing these agentic skills.

issue-import