mr-review-triage
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
subprocess.runto execute official CLI tools (git,gh,glab) andcurlfor repository management and API interaction. These executions are performed using list-based arguments without a shell, which prevents command injection vulnerabilities.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources (review comments).\n - Ingestion points:
ocr-pull-discussions.pyretrieves comments from GitLab MRs and GitHub PRs.\n - Boundary markers: The skill embeds invisible HTML comments (
<!-- triage-anchor: ... -->) in bot replies to track and deduplicate findings across sessions.\n - Capability inventory: The skill possesses capabilities to read and write review threads and manage labels using platform APIs via
ocr_github.pyandocr_gitlab.py.\n - Sanitization: External content is processed using regular expressions for structured 'fallback' notes, and outgoing payloads are serialized using
json.dumpsto ensure data integrity.\n- [EXTERNAL_DOWNLOADS]: The scripts interact with well-known service APIs (GitHub and GitLab). The destinations are eithergithub.comor user-defined GitLab instances derived from the local git remote, which is expected for the skill's functionality.
Audit Metadata