review-agent
Warn
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The file
LICENSE.txtincludes a copyright notice for 'OpenAI', which contradicts the provided author information ('hkuds'). This deceptive claim can mislead users into believing the skill is an official or trusted product from a well-known organization. - [INDIRECT_PROMPT_INJECTION]: The skill's design involves reading untrusted content which can contain malicious instructions. Ingestion points: The skill is instructed in
SKILL.mdto read theAGENTS.mdfile and code diffs from the repository. Boundary markers: There are no markers or instructions provided to separate the system prompt from potentially adversarial instructions in the repository files. Capability inventory: The skill usesgit merge-baseandgit difffor inspecting repository state. Sanitization: No sanitization or validation is applied to the content retrieved from the files being reviewed.
Audit Metadata