The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the external DeBank API, including human-readable transaction explanations and DeFi protocol details, which represents a potential surface for indirect prompt injection.
Ingestion points: External blockchain data is fetched from the DeBank API endpoint (pro-openapi.debank.com) within the tools/utils.py module and returned through the exports.py interface.
Boundary markers: Data retrieved from the API is passed to the agent without explicit delimiters or instructional boundary markers.
Capability inventory: The skill utilizes network operations via proxied_get and proxied_post from the platform's core.http_client and possesses transaction simulation capabilities.
Sanitization: No sanitization or content filtering is performed on the strings returned from the DeBank API before they are processed by the agent.
[DYNAMIC_EXECUTION]: The skill employs dynamic module loading to resolve its internal component structure.
Evidence: The exports.py file uses importlib.util.spec_from_file_location and loader.exec_module to load scripts from the tools/ directory. This implementation is used to maintain module isolation and prevents the skill's modules from shadowing standard Python libraries such as token or chain.

debank