vowel-client
Pass
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill documents an indirect prompt injection surface where the AI agent processes application context to determine its responses and actions.
- Ingestion points: Untrusted application state is ingested through
vowelInstance.updateContext()and theuseSyncContext()hook (SKILL.md). - Boundary markers: Documentation does not specify explicit delimiters or instructions for the agent to ignore potentially malicious content within the context data.
- Capability inventory: The agent can perform navigation using application adapters, execute custom actions (e.g.,
searchProducts), and interact with the DOM via the optionalautomationAdapter(SKILL.md). - Sanitization: There is no mention of sanitization or validation of the context data before it is presented to the LLM.
Audit Metadata