budge
Warn
Audited by Gen Agent Trust Hub on May 10, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill requires the inclusion of an external script from
https://www.budge.design/budge.iife.jsin the project layout, which executes vendor-provided JavaScript in the user's browser context.\n- [EXTERNAL_DOWNLOADS]: Runtime assets, including the IIFE script and audio feedback samples, are downloaded frombudge.design, which is not an established trusted domain.\n- [PROMPT_INJECTION]:\n - [Indirect Prompt Injection Surface]: The workflow involves an external script generating a text prompt for the user to paste back to the agent, allowing external influence over agent instructions (Ingestion: clipboard; Boundaries: none; Capabilities: file editing; Sanitization: none).\n
- [Metadata Poisoning]: The skill contains deceptive instructions in
website/AGENTS.mdandwebsite/package.jsonspecifying illegitimate versions of core packages (next@16.2.2,react@19.2.4) and commanding the agent to read local documentation paths for instructions, which is a known vector for redirecting agents to malicious instructions.
Audit Metadata