review
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local git commands (
git diff,git log) to retrieve information about code changes. It also generates 'One-shot prompts' that include suggested shell commands for users to run for verification purposes. - [DATA_EXFILTRATION]: The skill is designed to automatically post findings and comments to external GitHub Pull Requests if the environment supports the required tooling.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It ingests untrusted data from git diffs and commit logs, which are then passed to multiple subagents ('review hats'). A maliciously crafted diff could attempt to influence the logic of these subagents, suppress findings, or inject harmful shell commands into the generated 'One-shot prompts' provided to the user. Evidence: (1) Ingestion points:
git diff,git log, and various documentation files; (2) Boundary markers: Absent; (3) Capability inventory: command generation, subagent spawning, and PR commenting; (4) Sanitization: Not explicitly mentioned.
Audit Metadata