github-projects

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous gh CLI commands to enable the agent to manage GitHub issues and project boards.\n
  • Evidence: Bash snippets for gh issue, gh project, and gh api graphql are included for common tasks like creating issues, adding items to boards, and editing project fields.\n- [PROMPT_INJECTION]: The skill identifies a surface for potential indirect prompt injection through the ingestion of external repository and project metadata.\n
  • Ingestion points: Project IDs and settings are read from CLAUDE.md, and project titles and issue details are retrieved from the GitHub API via GraphQL and CLI commands.\n
  • Boundary markers: There are no explicit instructions for using delimiters or warnings when the agent interpolates or processes data from these external sources.\n
  • Capability inventory: The agent is granted the ability to create and edit issues and project items, which involves interpreting and using data from the aforementioned ingestion points.\n
  • Sanitization: The skill does not provide logic for sanitizing, validating, or escaping data retrieved from the repository or GitHub API.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 09:25 AM
Security Audit — agent-trust-hub — github-projects