agentic-workflows

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md instructs the agent to load multiple workflow prompt/skill files from github/gh-aw at runtime (e.g., .github/aw/*.md), which are outsider-authored repository content that becomes LLM-readable prose; this creates an indirect prompt-injection surface via the loaded markdown.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs at runtime to "Load these files from github/gh-aw" and then "follow it directly", so fetching content from the external repo github/gh-aw will control agent prompts and behavior.