chat-apps-ui-sdk
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with user-provided instructions for app development tasks.
- Ingestion points: User prompts triggered by keywords in
SKILL.mdsuch as 'chat app' or 'render UI in chat'. - Boundary markers: The skill does not explicitly define delimiters for user-provided context.
- Capability inventory: The skill provides templates for creating tools with network and file-access capabilities, documented in
references/sec-declare-csp-allowlist.mdandreferences/bridge-call-tools-app-visibility.md. - Sanitization: Security is addressed via a set of rules (e.g.,
sec-enforce-server-side-auth.md) that guide the agent to generate secure implementations. - [CREDENTIALS_UNSAFE]: Truncated placeholder keys appear in documentation as examples of insecure coding.
- Evidence:
references/sec-no-secrets-in-payloads.mdandreferences/state-no-secrets-in-state.mdcontain example strings like 'sk_live_51H...' and 'atk_live_8Q2x...'. These are non-functional and clearly marked as incorrect practices.
Audit Metadata