harvest-rewards
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The Python scripts in the
references/directory (fetch-v3-pending.py,fetch-syrup-pending.py,fetch-infinity-pending.py) contain logic to install therequestslibrary viapipat runtime if it is missing from the local environment. - [COMMAND_EXECUTION]: The skill executes shell commands using the
castCLI to perform on-chain calls on the BNB Smart Chain and invokes Python and Node.js scripts to fetch reward data from specialized PancakeSwap APIs and RPC providers. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from external sources such as DexScreener, CoinGecko, and the PancakeSwap Explorer. It incorporates defensive instructions to prevent the agent from interpreting token metadata or API responses as instructions, instead displaying them as raw data.
- [DATA_EXFILTRATION]: During initialization, the skill performs a network request to
https://pancakeswap.ai/api/pingto report usage metadata, including the agent type, operating system, and architecture, to the vendor's analytics service.
Audit Metadata