The Agent Skills Directory

[SAFE]: The skill performs its intended function of staking reward optimization using standard Python libraries and well-known cryptocurrency data providers. Analysis of the scripts and instructions reveals no malicious behavior.
[EXTERNAL_DOWNLOADS]: The skill fetches staking pool data and asset prices from DeFiLlama (yields.llama.fi) and CoinGecko (api.coingecko.com). These are established and well-known services in the blockchain ecosystem, and the references are documented neutrally.
[COMMAND_EXECUTION]: The skill runs local Python scripts (staking_optimizer.py) to process data and generate reports. These operations are restricted by the platform's tool configuration.
[DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive local data access (e.g., credentials, private keys) or exfiltration to untrusted domains was identified. The skill maintains a local cache file (~/.staking_optimizer_cache.json) for public API data.
[SAFE]: Analysis for indirect prompt injection surfaces shows that while the skill ingests external data, it processes it safely.
Ingestion points: External staking data is ingested via scripts/staking_fetcher.py from DeFiLlama and CoinGecko APIs.
Boundary markers: No explicit boundary markers or instruction-ignoring delimiters are used for the external data.
Capability inventory: The skill has access to file system operations (Read, Write, Edit) and a scoped Bash tool as defined in SKILL.md.
Sanitization: External data is parsed as structured JSON and processed as specific numerical or string fields before being formatted for output.

optimizing-staking-rewards