actions-optimiser

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

• Third-party content exposure detected (high risk: 0.95). The skill’s required workflow fetches and decodes YAML from public GitHub repositories at runtime (via gh search code and gh api .../contents/...), so outsider-authored free text (workflow files written by other users) is ingested into the agent’s LLM context after base64 decoding.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill issues a runtime GitHub API call (e.g., "gh api repos/Sofiane-Truman/cloud-foundation-fabric/contents/.github/workflows/linting.yml") to fetch workflow YAML which is then decoded and used to drive suggested edits, so the external content directly controls the agent's behavior.