The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (The Graph subgraphs and DeFiLlama API) which is then processed and formatted for the agent/user. This creates a surface where malicious instructions hidden in external data could potentially influence the agent.
Ingestion points: scripts/pool_fetcher.py performs HTTP requests to fetch pool data using the requests library.
Capability inventory: The skill has file read/write permissions and can execute bash commands via the pool_analyzer.py script.
Boundary markers: The skill does not implement explicit delimiters or warnings when handling data fetched from APIs.
Sanitization: While data is parsed as JSON, there is no evidence of specific sanitization or filtering of string values (like token symbols or names) which could contain adversarial prompts.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions for installing a standard Python dependency.
Evidence: references/errors.md contains the command pip install requests as a solution for missing library errors.
Note: The requests library is a well-known and trusted package in the Python ecosystem, used here to interact with established DeFi data providers.

analyzing-liquidity-pools