4-step-program

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

• Third-party content exposure detected (high risk: 0.90). The workflow explicitly instructs the agent to fetch and parse user-generated GitHub content (e.g., "gh issue view " and line-by-line extraction of requirements in SKILL.md) and to act on those findings (drive delegation, verification, and PR actions), so untrusted third-party issue/PR text can materially influence tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires fetching GitHub issue text at runtime (e.g., via "gh issue view ") and using URLs like https://github.com/owner/repo/issues/100 to extract ALL requirements that directly control delegation prompts and agent behavior, making this a runtime external dependency that controls prompts.