The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's workflow involves executing local shell commands such as rg, node, and ls to find, process, and verify session files. This is the intended behavior for managing local session logs.
[DATA_EXPOSURE]: The skill reads and modifies files located in ~/.codex/sessions and ~/.codex/session_index.jsonl. This access is necessary for the task of cleaning session history and no data is exfiltrated to external systems.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted session logs through its Node.js script.
Ingestion points: The scripts/scrub-codex-session-images.mjs script reads data from .jsonl files in ~/.codex/sessions.
Boundary markers: Absent, as data is processed as structured JSON rather than natural language instructions.
Capability inventory: The script performs file read/write operations on the local disk but lacks network access or dynamic code execution capabilities.
Sanitization: The script uses strict JSON.parse and JSON.stringify logic, replacing image-related fields with a hardcoded placeholder string instead of executing or interpolating the content.

codex-session-image-scrubber