eval-integrity
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill audits external repositories by reading and analyzing source code, CI logs, and documentation across seven integrity dimensions.
- Ingestion points: The skill reads arbitrary content from a user-provided benchmark repository path using
rgandgitcommands as specified inSKILL.mdandpatterns/dimension-prompts.md. - Boundary markers: The prompts provided to the subagents do not include explicit delimiters or 'ignore embedded instructions' warnings, which could allow a malicious repository to influence the auditor's findings via hidden text.
- Capability inventory: The skill's operations are restricted to read-only search and listing tools (
rg,git,gh). It explicitly states a 'read-only' posture and does not include instructions for code execution or file modification. - Sanitization: There is no logic provided to sanitize or filter external content before it is processed by the agent or subagents.
Audit Metadata