visual-recap

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

• Third-party content exposure detected (high risk: 0.75). The required workflow for /visual-recap ingests the PR/commit diff and thread context (outsider-authored code/comments from other contributors) and then uses that content to generate the recap’s prose and structured blocks, which are fed into the agent’s LLM context for summarization (indirect prompt-injection risk).

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 the Plan MCP block catalog and reconnecting to the hosted MCP at https://plan.agent-native.com at runtime (via get-plan-blocks or npx reconnect), which supplies the authoritative block vocabulary the agent must read and that directly controls how the agent composes plan content.