ui-continuity
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user feedback and external design specifications which could potentially contain malicious instructions.\n
- Ingestion points: User feedback strings and external authority sources (registries, specs) are ingested as described in
SKILL.md.\n - Boundary markers: The skill mandates the use of a "Scoped Feedback Protocol" (in
references/feedback-protocol.md) and a "Change Budget" (inreferences/continuity-contract.md) to define strict boundaries for agent behavior.\n - Capability inventory: The skill primarily generates structured text reports and diffs. It mentions handing off tasks to other rendering skills (e.g.,
huashu-design,Stitch,Pencil), but the provided files do not contain any network operations, file-write capabilities, or subprocess executions.\n - Sanitization: The workflow includes a mandatory baseline-to-candidate diffing process and regression checklists to ensure the agent does not deviate from the approved structure.
Audit Metadata