lov-codex-thread-usage
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE_EXFILTRATION]: The skill accesses local Codex data files, including SQLite databases and rollout JSONL logs, to calculate token metrics. These operations are performed in a read-only manner on local paths (typically within the user's home directory). No network capabilities or exfiltration vectors were identified in the script logic.
- [CREDENTIALS_UNSAFE]: The profile management script
scripts/profile_store.pycontains logic to explicitly block the storage of sensitive keys such as 'token', 'secret', 'password', 'credential', and 'api_key', ensuring that secrets are not accidentally persisted in the shared user profile context. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from local Codex rollout files. However, the analysis is limited to extracting token counts and metadata via structured JSON parsing and regex validation. The skill does not interpret or execute the content of the conversation messages as instructions.
- [COMMAND_EXECUTION]: The skill executes local Python scripts provided within its own directory to perform diagnostics. These scripts are self-contained and do not accept arbitrary shell commands or untrusted external inputs for execution. It relies on standard environment variables for configuration.
Audit Metadata