code-reviewers
Pass
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface through the processing of untrusted repository code diffs.
- Ingestion points: Sub-agents (Seymour-Bugs and Stack-Tracy) are instructed to read and inspect code diffs directly.
- Boundary markers: The prompt template provided for the reviewers lacks specific delimiters or explicit instructions to treat the diff content as data only, potentially allowing embedded malicious instructions to influence agent behavior.
- Capability inventory: The orchestrator and sub-agents use the
AgentandSendMessagetools; while these are primarily for coordination, they provide a communication path for exploited sub-agents. - Sanitization: There are no instructions or mechanisms defined to sanitize or escape the content of the diffs before they are presented to the AI models.
Audit Metadata