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 conversation context (outsider-authored code/PR text) to generate the recap, which is then fed into the agent’s LLM context for summarization and block construction—an indirect prompt-injection risk.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs running the remote CLI (e.g., npx @agent-native/core@latest) at runtime to fetch the Plan block catalog and connect to the Plan MCP origin https://plan.agent-native.com, which executes remote code and supplies the authoritative block/schema that directly controls what the agent writes, so this is a runtime external dependency tied to agent behavior.