cicd-pipeline-builder

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow includes “Source Refresh” steps that fetch external web content (e.g., DORA research updates and SLSA spec changes via https://dora.dev/research/ and https://slsa.dev/spec/), which can be ingested as free-form text into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The Research Checklist contains explicit runtime invocations of adn_skills that fetch external content (e.g., adn_skills("distill_from_wikipedia", topic="Continuous delivery") and adn_skills("import_from_github", repository="org/repo", ...") — i.e. pulling from sources like https://en.wikipedia.org/wiki/Continuous_delivery and https://github.com/org/repo) which would be loaded at runtime and can directly influence prompts or bootstrap remote code, so this is a runtime external dependency that can control agent behavior.