codeguard-git-branch
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions emphasize read-only assessment of Git metadata and require user authorization for all write actions such as branch creation, merging, or rebasing. It operates on local repository data using standard Git tooling.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from branch names and repository logs which could be controlled by an attacker, but it mitigates risks via strict validation requirements. 1. Ingestion points: Reads branch lists and commit logs via
git branch -randgit log(SKILL.md). 2. Boundary markers: Includes instructions to identify the model and follow 'Capability Boundaries' and 'Evidenceized Workflow' (SKILL.md). 3. Capability inventory: Includes the ability to execute git write commands for branching, merging, and tagging in SKILL.md. 4. Sanitization: Employs structural pattern matching against defined branching models to validate inputs (SKILL.md).
Audit Metadata