versa-deploy
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the 'versa' and 'bankr' skills from the BankrBot GitHub repository.
- [COMMAND_EXECUTION]: Utilizes the 'bankr' utility to sign and execute smart contract functions (registerAgent, withdrawFees, deactivateAgent) on the Base mainnet (chain ID 8453).
- [DATA_EXFILTRATION]: Transmits configuration data, including the user-defined 'secretPhrase', 'defensePrompt', and transaction hashes, to the backend service at versa-production.up.railway.app.
- [PROMPT_INJECTION]:
- Contains 'CRITICAL RULES' intended to override default agent behavior to enforce platform-specific constraints, such as secret phrase length and treasury withdrawal locks.
- The 'contextual' mode creates a significant surface for indirect prompt injection by exposing the agent to all prior challenge attempts from various users.
- Ingestion points: Challenge attempts from all players are aggregated and provided as context to the agent in 'contextual' mode, as described in references/parameters.md.
- Boundary markers: The instructions do not define delimiters or specific warnings to the agent to treat historical challenge data as untrusted or to ignore instructions embedded within those messages.
- Capability inventory: The skill possesses capabilities to perform onchain financial transactions and communicate with an external API.
- Sanitization: The skill instructions do not specify any sanitization, filtering, or normalization of the historical challenge data before it is presented to the agent.
Audit Metadata