The Agent Skills Directory

[SAFE]: The skill operates as a requirements-gathering tool, using Socratic questioning to clarify user intent. It manages project-specific artifacts in a local .rcs/ directory and does not access sensitive credentials or perform unauthorized network operations.\n- [COMMAND_EXECUTION]: The skill uses local CLI tools like rcs and explore for interactive questioning and codebase analysis. These are system-integrated tools used for their intended purpose of providing structured context to the agent.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests user input to generate specs for other autonomous skills. It mitigates this risk by providing semantic pressure through iterative questioning and maintaining a summary gate for large inputs. Ingestion points: Initial task arguments and Socratic interview answers. Boundary markers: Prompt-safe context summary gates and structured artifact generation. Capability inventory: Codebase exploration, state persistence, and handoff to planning/execution skills. Sanitization: Semantic intent validation through Socratic clarification and assumption-testing.

deep-interview