github-workflow-automation

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

• Third-party content exposure detected (high risk: 0.95). Outsider-authored free text from GitHub issue/PR comments and bodies (e.g., github.event.comment.body, context.payload.issue.body, and PR diffs from the repo) is embedded into the LLM prompts at runtime via the workflow’s actions/github-script step(s), which then calls the AI.

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

• Potentially malicious external URL detected (high risk: 0.90). The workflows call the external GitHub Action actions/github-script@v7 (e.g., SKILL.md:58), which is fetched and executed at runtime and runs JavaScript used to build and send prompts to AI, so it is a runtime dependency that executes remote code.