Skills
Navigate
skills/adonis0123/adonis-skills/kimi-computer-use/Gen Agent Trust Hub

kimi-computer-use

Pass

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from running macOS applications via the get_app_state tool, creating an attack surface where malicious content in an app UI could influence agent behavior.
  • Ingestion points: Untrusted data enters the agent context through the output of the get_app_state tool (with mode: "ax", image, or full) when inspecting application states, as described in SKILL.md.
  • Boundary markers: The SKILL.md file defines a "Safety boundary" section requiring user confirmation for consequential actions and providing specific rules for UI verification (e.g., marking zero-call checks as UNVERIFIED).
  • Capability inventory: The skill has the capability to execute the kimi-cu CLI for management tasks and modify host configuration files (like mcp.json) as detailed in SKILL.md and references/host-configs.md.
  • Sanitization: The instructions do not specify any explicit sanitization or escaping of the application UI data before it is analyzed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 31, 2026, 02:41 PM