bitflow-hodlmm-withdraw
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The skill accesses sensitive local files within the agent's environment to retrieve private keys and session data.
- Evidence: The script bitflow-hodlmm-withdraw.ts reads from ~/.aibtc/wallets.json, ~/.aibtc/config.json, ~/.aibtc/wallets/*/keystore.json, and ~/.aibtc/sessions/.session-key to resolve signers and decrypt mnemonics.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external APIs to calculate transaction parameters, which could be manipulated by a compromised data source to influence financial outcomes.
- Ingestion points: Bitflow API endpoints in bitflow-hodlmm-withdraw.ts (functions getPool, getBins, getUserBins).
- Boundary markers: Not utilized in prompt interpolation.
- Capability inventory: Blockchain transaction broadcasting (broadcastTransaction in bitflow-hodlmm-withdraw.ts).
- Sanitization: Performs basic type parsing and validation but lacks protection against adversarial logic manipulation from the API source.
- [EXTERNAL_DOWNLOADS]: The skill makes network requests to external services to retrieve market and blockchain state.
- Evidence: Fetches data from https://api.hiro.so and https://bff.bitflowapis.finance.
- [COMMAND_EXECUTION]: The skill documentation and AGENT.md instruct the agent to execute the TypeScript script using the bun runtime to perform blockchain operations.
- Evidence: Execution of commands like bun run bitflow-hodlmm-withdraw/bitflow-hodlmm-withdraw.ts run --confirm=EXIT.
Audit Metadata