The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Queries public blockchain data from well-known services (blockchain.info and blockstream.info). These network operations are necessary for the skill's functionality and do not involve downloading or executing external code.\n- [COMMAND_EXECUTION]: The script processes user-supplied Bitcoin addresses via command-line arguments. It implements robust input validation using regular expressions to ensure inputs conform to expected cryptocurrency address formats.\n- [DATA_EXFILTRATION]: The skill does not access sensitive local files, credentials, or environment variables. Its network activity is restricted to retrieving public blockchain information.\n- [PROMPT_INJECTION]: The skill processes data from external APIs (blockchain explorers). \n
Ingestion points: scripts/agent.py (API responses from blockchain.info and blockstream.info).\n
Boundary markers: None.\n
Capability inventory: Includes network read access and console output; no capabilities for file system writes or dynamic code execution were found.\n
Sanitization: Input validation for wallet addresses is present. The structured nature of the retrieved blockchain data (hashes, amounts) poses a negligible risk for indirect prompt injection.\n- [SAFE]: The automated detection of a Trojan in scripts/agent.py is a false positive. The signature match is triggered by the hardcoded 'Indicators of Compromise' (known ransomware wallet addresses) which are legitimately included for attribution analysis.

analyzing-ransomware-payment-wallets