agreement
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill manages a workflow for processing external documents, such as signed agreements, Statements of Work (SOWs), and amendments, which may contain untrusted data.
- Ingestion points: Content is ingested from external agreement and contract documents during initialization and amendment recording, as defined in
SKILL.mdandhost-contract.json. - Boundary markers: The skill utilizes structured templates in
templates/AGREEMENT.mdwith explicit<!-- agreement:start -->and<!-- agreement:end -->delimiters to scope the agreement context within thePROJECT.mdfile and isolate it from other project data. - Capability inventory: The skill utilizes
Bash,Write,Edit, andReadtools for document maintenance and resolving workspace context. - Sanitization: A mandatory "native mutation sequence" is defined in
SKILL.md(Inspect → Prepare → Verify preparation → Apply → Verify result → Complete), which enforces a multi-step manual verification process requiring the agent to re-read and validate all prepared changes before committing them to persistent storage.
Audit Metadata