The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): The skill is designed to surface internal system states and data flows through instrumentation. \n
Evidence: Steps 2 and 6 focus on revealing hidden states, transitions, and logic via emitters and proxies. \n
Mitigation: Step 7 includes a specific check (process.env.NODE_ENV !== 'development') to prevent the visualization routes from being accessible in production environments.\n- [Dynamic Execution] (LOW): The skill instructs the agent to generate new application routes and instrumentation scripts. \n
Evidence: Steps 7 and 8 describe creating app/__dev/ routes and src/lib/ files. \n
Context: The generation is template-based and for developer tooling; the skill includes explicit 'EXPLANATORY-PLAYGROUND' markers and cleanup steps to facilitate the removal of generated code.\n- [Indirect Prompt Injection] (INFO): The skill creates an attack surface by ingesting application data for visualization. \n
Ingestion points: System state captured via EventEmitter and Proxy objects in Step 6. \n
Boundary markers: None specified for the data being visualized. \n
Capability inventory: Visualization (display-only) within a React component. \n
Sanitization: None specified; however, the impact is limited to display-only within a development route.\n- [Unverifiable Dependencies & Remote Code Execution] (INFO): Recommends the installation of external UI libraries. \n
Evidence: Step 3 suggests using react-flow and recharts for building the visualization interfaces.

explanatory-playground