snap-forge

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and interprets GitHub issue comments as part of its required workflow ("If input is a GitHub issue, fetch comments via gh issue view <number> --comments") and uses plan comment contents (e.g., "Durable Decisions" / "Phase") to drive constraints and behavior, which exposes the agent to untrusted, user-generated third-party content that can change its actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches GitHub issue comments at runtime using "gh issue view --comments", which pulls content from GitHub (e.g., https://api.github.com) and uses plan-comments (## Durable Decisions / ## Phase) to directly control the agent's behavior, so this is a runtime external dependency that controls prompts/instructions.