The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill guides the agent to write and execute ad-hoc scripts for complex analytical tasks such as transaction aggregation, date-range filtering, and multi-step blockchain queries. It emphasizes using the standard library and provides a regulated process for installing third-party packages if necessary.
[EXTERNAL_DOWNLOADS]: The skill retrieves blockchain data and metadata from official Blockscout endpoints, including the MCP server and the Chainscout registry. These are legitimate resources provided by the skill author for its intended functionality.
[INDIRECT_PROMPT_INJECTION]: The skill includes a dedicated security section that identifies blockchain data as untrusted and potentially adversarial. It provides explicit instructions for the agent to sanitize and summarize this data.
Ingestion points: External data from blockchain explorers and indexers via the Blockscout MCP REST API.
Boundary markers: Instructions are included to separate user instructions from retrieved data and to avoid interpreting data as commands.
Capability inventory: Capabilities include read-only contract calls, generalized API queries, and runtime script execution.
Sanitization: The agent is instructed to summarize or sanitize API response data before processing or outputting it.
[COMMAND_EXECUTION]: The agent executes its own generated scripts to handle deterministic data processing. The skill mandates specific headers (User-Agent) and retry logic to ensure these operations are performed correctly and predictably.

blockscout-analysis