github-agentic-workflows

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The SKILL instructs the agent to read the local references (SKILL.md Step 2) which in turn (references/examples.md and references/authoring.md) tell agents to use external raw.githubusercontent.com URLs (e.g., "Create a workflow ... using https://raw.githubusercontent.com/github/gh-aw/main/create.md"), meaning the agent may fetch and interpret public third‑party content that can change its actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill includes a runtime installation command that fetches and executes remote code (curl -sL https://raw.githubusercontent.com/github/gh-aw/main/install-gh-aw.sh | bash in references/troubleshooting.md), which directly runs external code and is presented as a required fallback for installing the gh-aw tooling.