tavus-cvi-ui
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill utilizes
npx @tavus/cvi-ui@latestto initialize projects and add components. These are vendor-provided tools from Tavus Engineering used for project scaffolding. - [CREDENTIALS_UNSAFE]: Instructions include setting up a
.envfile for API keys and persona IDs. The skill explicitly recommends server-side API calls to keep theTAVUS_API_KEYsecure, aligning with standard security best practices for web development. - [INDIRECT_PROMPT_INJECTION]: The
useAppMessagehook ingests utterance data from conversations. While this processes external content, the skill does not possess capabilities that would allow this data to trigger unauthorized actions or privilege escalation. - Ingestion points: The
useAppMessagehook inSKILL.md(e.g.,event.properties.content). - Boundary markers: None explicitly defined in the UI code examples.
- Capability inventory: The skill provides UI components and hooks; no shell execution, file system modification, or dangerous network operations are present.
- Sanitization: Not explicitly shown in the basic code examples, but no sensitive sinks are reached.
Audit Metadata