w-v0
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill drafts prompts using untrusted content from the local environment (current file, selection, and stack). * Ingestion points: The skill processes local file content and stack context as described in SKILL.md. * Boundary markers: Absent; there are no delimiters or markers to separate the untrusted data from the drafting instructions. * Capability inventory: The skill has the capability to open external URLs via the browser as defined in SKILL.md. * Sanitization: Present; the skill relies on a human-in-the-loop approval step ('wait for OPEN') and explicit instructions to the agent to exclude secrets and internal paths.
- [DATA_EXFILTRATION]: The skill transmits local context to the external domain w-v0.dev via URL parameters. While this is the primary function of the skill, it involves sending user-derived data to a third-party service. The skill includes instructions to mitigate the risk of exfiltrating sensitive credentials or repository-internal paths.
Audit Metadata