bitget-wallet
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from the Bitget Wallet API, including token narratives, AI-generated summaries, and market metadata which are presented to the user.
- Ingestion points: API response processing in
bitget-wallet-agent-api.pyandx402_pay.py. - Boundary markers: Absent in the documentation for presenting external API content.
- Capability inventory: Transaction signing and broadcasting capabilities across several scripts including
order_sign.pyandbitget-wallet-agent-api.py. - Sanitization: No explicit sanitization or filtering of strings returned from the API is performed before interpolation into agent responses.
- [EXTERNAL_DOWNLOADS]: The skill makes network requests to the vendor-controlled API at
copenapi.bgwapi.iofor its primary operations and communicates with third-party resource servers as part of the x402 payment protocol. - [DYNAMIC_EXECUTION]: Scripts such as
order_make_sign_send.py,social_order_make_sign_send.py, andsocial_transfer_make_sign_send.pyuse dynamic loading viaimportlibto load the local API client module. - [PROMPT_INJECTION]: The
SKILL.mdcontains defensive instructions that attempt to override agent behavior regarding transparency, specifically instructing the agent never to read or explain the source code of the signing scripts or reveal stored application secrets.
Audit Metadata