The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions in implementation.md direct the agent to load API credentials from a sensitive file path: {baseDir}/config/crypto-apis.env. This file is used to store authentication tokens for blockchain explorers and RPC services.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill performs network operations to external services to retrieve gas and price data.
Network operations: Fetches data from api.coingecko.com, eth.llamarpc.com, api.etherscan.io, and other chain-specific RPC/explorer endpoints. These are well-known services used for their intended purpose in this skill.
File access: The scripts gas_fetcher.py and pattern_analyzer.py write data to the user's home directory (e.g., ~/.gas_cache_{chain}.json and ~/.gas_history.json) to maintain performance and historical context.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external, untrusted sources which creates a potential surface for indirect prompt injection.
Ingestion points: API responses from Etherscan, CoinGecko, and various RPC providers are processed in gas_fetcher.py and cost_estimator.py.
Boundary markers: The prompt instructions in SKILL.md lack explicit boundary markers or warnings to ignore instructions that might be embedded in the retrieved blockchain or price data.
Capability inventory: The skill utilizes the Bash tool to execute Python scripts that process this data. However, it does not currently use the data to construct shell commands or execute dynamic code.
Sanitization: The data is parsed using standard JSON libraries (requests.json() and json.load()), which provides basic structure validation.

optimizing-gas-fees