The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/codex_skill_admin.py uses subprocess.Popen and subprocess.run to interact with the local codex CLI tool. This is used to manage a temporary application server for communication and to verify skill visibility in model prompts. These operations are restricted to the local environment and the official platform binary.
[DATA_EXFILTRATION]: The skill accesses local session history files located in ~/.codex/sessions and related directories to audit which skills have been used. This data is processed entirely locally to generate usage statistics; no external network requests or exfiltration attempts were found. Network activity is confined to 127.0.0.1.
[EXTERNAL_DOWNLOADS]: The skill does not download any external dependencies, scripts, or binary payloads. It relies exclusively on the Python standard library and the pre-installed codex CLI.
[PROMPT_INJECTION]: Analysis of the instructions and metadata revealed no attempts to override agent safety guidelines, bypass restrictions, or leak system prompts. The skill's instructions focus on administrative tasks and usage auditing.

codex-skill-admin