chatcut-plugin-basics-kimi-work
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill handles a sensitive
editor-boot-tokenused for session handoff and auto-authentication within an in-app browser. Critically, it includes a mandatory security rule (Rule 4 under Editor Handoff) requiring the agent to strip the token from any URL before displaying it to the user in text or Markdown, mitigating the risk of accidental credential exposure. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external ChatCut project states, including timeline items, transcription text, and asset metadata. Because these fields can contain user-generated or third-party content, they represent an indirect prompt injection surface. The skill mitigates this by requiring explicit verification via the connector (
read_project) rather than relying on potentially poisoned browser-side visual states.
Audit Metadata