The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection as it is designed to ingest and process arbitrary data from various blockchains.
Ingestion points: Data retrieved from 159 tools across 100+ chains, including NFT metadata, transaction receipts, logs, and trace outputs (SKILL.md).
Boundary markers: The instructions do not specify any delimiters or special handling to separate external blockchain data from the agent's control context.
Capability inventory: The skill includes administrative capabilities such as update_allowlist, create_app, and update_app which could be targets for malicious instructions embedded in blockchain data.
Sanitization: No mention of content sanitization or validation of the strings retrieved from on-chain sources.
[SAFE]: All network references and configuration instructions point to the vendor's official infrastructure (mcp.alchemy.com, alchemy.com) and GitHub organization (alchemyplatform). These are standard operations for a hosted MCP service.

alchemy-mcp