gitlab-mr
Warn
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill contains shell commands designed to extract a GitLab private token from the local filesystem at 'C:\Users<user>\AppData\Local\glab-cli\config.yml'.
- [COMMAND_EXECUTION]: The skill utilizes several command-line tools, including 'glab', 'git', 'curl', and 'awk', to perform operations such as creating merge requests, checking for conflicts, and interacting with the GitLab API.
- [DYNAMIC_EXECUTION]: To process JSON data from API responses, the skill pipes output into inline Python scripts ('python -c') to extract specific fields like user IDs and merge status.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from GitLab merge requests, including descriptions and comments, which could potentially contain untrusted instructions.
- Ingestion points: Data is fetched from the GitLab API via 'curl' and 'glab mr view' (found in SKILL.md and references/local-environment.md).
- Boundary markers: No specific delimiters or safety instructions are defined for the handling of ingested merge request content.
- Capability inventory: The skill possesses the ability to execute shell commands, perform write operations via the GitLab API (merge, approve, close, comment), and read files from the local filesystem.
- Sanitization: While structured JSON parsing is used, there is no explicit sanitization or filtering of natural language content before it is processed by the agent.
Audit Metadata