The Agent Skills Directory

[COMMAND_EXECUTION]: Uses the subprocess module to invoke the codex command-line utility for generating qualitative data summaries and narrative insights. This is an expected part of the tool's diagnostic and reporting functionality. It also manages its own configuration and authentication files in ~/.codex to facilitate execution isolation.\n- [PROMPT_INJECTION]: Processes local session logs containing user-generated text and includes them in prompts for an LLM. This creates a surface for indirect prompt injection, which is addressed through internal redaction filters for secrets and HTML escaping during the report rendering phase.

codex-insights