diagnose-ocr-ci
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The script reads GitLab authentication tokens from the local configuration file at
~/.config/glab-cli/config.yml. These tokens are used to authorize API requests to GitLab instances for fetching job details and artifacts. - [COMMAND_EXECUTION]: The skill requires the user to execute a local Python script using
uv runto perform the analysis of CI and local runs. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and parses external data from GitLab job traces and artifacts. 1. Ingestion points: GitLab API endpoints for jobs, artifacts, and traces, as well as local session files. 2. Boundary markers: No specific boundary markers are used to isolate untrusted log content during regex-based signature matching. 3. Capability inventory: The script performs network reads, reads local configuration and session data, and writes analysis output to the temporary directory
/tmp/diagnose-ocr-ci/. 4. Sanitization: ANSI escape sequences are removed from log data before processing using thestrip_ansifunction.
Audit Metadata