The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes platform-specific tools including omx question and omx explore via shell execution to interact with the user and gather codebase facts.
[DATA_EXFILTRATION]: Manages and stores session state, interview transcripts, and execution specifications within the local .omx/ project directory. These operations are restricted to the local file system for metadata persistence.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an ingestion surface for user-provided descriptions and codebase data which are then compiled into specifications for downstream agents.
Ingestion points: Captures untrusted data through user answers in the omx question tool and codebase information retrieved via the explore command.
Boundary markers: The skill instructions do not specify explicit delimiters or markers to separate user-provided content from agent instructions in the final Markdown artifacts.
Capability inventory: Possesses capabilities to read from the codebase (explore), write state and artifacts (state_write), and interact with the user via structured questioning.
Sanitization: No explicit sanitization or escaping mechanisms are described for user input before it is written to the .omx/specs/ directory.

deep-interview