issue-driven-github-flow
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill utilizes official GitHub CLI (gh) and Git commands for repository management, keeping operations within the intended developer workspace.
- [SAFE]: Secret management guidance is integrated, explicitly instructing agents to avoid committing credentials and providing alternatives like environment variables and GitHub Secrets.
- [SAFE]: Repository mutation operations, such as applying branch protection or merging PRs, are gated by requirements for human oversight or explicit consent.
- [PROMPT_INJECTION]: The skill processes untrusted content from GitHub issues and pull requests, creating a surface for indirect prompt injection attacks. 1. Ingestion points: Data enters the context via gh issue view, gh pr view, and git diff outputs used by sub-agents. 2. Boundary markers: The skill uses quoted heredocs (<<'EOF') when passing content to shell tools, effectively preventing shell-level injection. 3. Capability inventory: Agents have capabilities to write to the repository, create branches, and post comments on GitHub issues and PRs. 4. Sanitization: Security is enhanced through role separation (Planning, Review, Implementation) and enforcing a maximum of three rounds for agent-driven loops to prevent autonomous abuse.
Audit Metadata