hai-rewrite-doc
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 documents that may contain malicious instructions or commands embedded by an attacker.\n
- Ingestion points: The skill takes a user-provided 'old document' as raw material for rewriting (SKILL.md, Workflow Step 3).\n
- Boundary markers: There are no instructions to use delimiters or specific safety markers to ignore embedded instructions within the source document content.\n
- Capability inventory: The agent is explicitly directed to verify claims against 'runnable commands' and 'code' cited in the document (SKILL.md, Core Principle), and has capabilities to perform file-write operations ('Edit the original in place'). Executing commands found in an untrusted source document is a high-risk activity.\n
- Sanitization: No sanitization, escaping, or validation of the document's content is performed before the agent treats it as 'evidence' for verification and execution.
Audit Metadata