The Agent Skills Directory

[COMMAND_EXECUTION]: The "Prototype and Compare" phase (Phase 4) explicitly directs the agent to "Build minimal prototypes" and "Run the comparison." This process involves writing functional code to the filesystem and executing it to measure benchmarks or ergonomics. While the skill limits prototypes to a small scope (50-200 lines), it still constitutes the execution of dynamically generated code based on external and internal inputs.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it uses external data to influence the agent's planning and execution logic.
Ingestion points: The skill performs "substantive research" using web search and documentation lookups to find paradigms and libraries.
Boundary markers: The instructions lack specific delimiters or "ignore embedded instructions" warnings for the agent when processing external research content.
Capability inventory: The skill possesses file-system write access (to create exploration artifacts and prototypes) and command execution capabilities (to run benchmarks).
Sanitization: No validation or sanitization of external research data is required before it is used to generate the "Solution Map" or "Analysis" artifacts.
[EXTERNAL_DOWNLOADS]: The skill encourages the agent to research technology paradigms using "web search" and "documentation lookups," which involves fetching data from non-whitelisted external domains. This is a standard functional requirement for technical research but presents a surface for untrusted data ingestion.

solution-explorer