hai-audit-docs
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user documents and compares them against source code or configurations. This represents a potential surface for indirect prompt injection, where instructions embedded within the analyzed documents could attempt to influence the agent's auditing logic. However, the risk is minimal as the skill's capabilities are restricted to document analysis and local file modification (repairs) without network or shell execution access.
- Ingestion points: The skill reads user-specified document sets, source code, and configuration files as evidence for its audit.
- Boundary markers: The instructions do not define explicit delimiters for untrusted content, though they mandate the use of structured output templates.
- Capability inventory: The skill is authorized to perform "local repairs" (file writes) to documents when specifically requested by the user. It does not perform network operations or execute shell commands.
- Sanitization: No specific sanitization or filtering of document content is described in the workflow.
- [SAFE]: No malicious patterns, credential harvesting, or unauthorized network activity were detected. The skill follows best practices by scoping its execution to user-defined targets and providing clear boundaries between auditing and editing functions.
Audit Metadata