create-merge-request-changelog
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted git commit messages, unified diffs, and GitLab merge request content, which are interpolated into an AI prompt for categorization and summary generation. This creates a surface for indirect prompt injection where malicious instructions in a commit could potentially influence the AI output. \n
- Ingestion points:
scripts/analyze_git_changes.py(extracts git logs and diffs) andscripts/fetch_gitlab_mr.py(fetches GitLab MR metadata and content). \n - Boundary markers: The primary analysis prompt in
references/analysis_prompts.mduses XML-like delimiters (e.g.,<commits>,<diff>,<file_list>) to separate untrusted data from system instructions. \n - Capability inventory: The skill can read from local git repositories, access the GitLab API, and write analysis results to the local filesystem. It does not have arbitrary command execution or external network exfiltration capabilities. \n
- Sanitization: No explicit sanitization or escaping of the ingested git data is performed before it is passed to the AI model.
Audit Metadata