The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation references the download and installation of official Subsquid packages and templates from well-known sources.
Fetches the @subsquid/cli global package via npm.
References official GitHub repositories for documentation and sample code under the subsquid and subsquid-labs organizations.
Describes using sqd init with a template URL to scaffold new projects from GitHub repositories.
[COMMAND_EXECUTION]: The skill details the use of CLI tools and npx commands for generating code and running indexers.
Instructions include running sqd run, npx squid-typeorm-codegen, npx squid-evm-typegen, and npx squid-graphql-server.
These commands are standard for the documented development workflow and operate within the project context.
[INDIRECT_PROMPT_INJECTION]: As a blockchain indexing toolkit, the skill defines a surface area for indirect prompt injection through the ingestion of arbitrary chain data.
Ingestion points: Data enters the agent's context through processor.run() in references/core-evm-processor.md, which receives block headers, logs, and transactions from SQD Network gateways or RPC endpoints.
Boundary markers: The documentation demonstrates the use of typegen for decoding data (e.g., usdcAbi.events.Transfer.decode(log)), which provides structure but does not act as a security boundary against injection attacks hidden within blockchain transaction data or event logs.
Capability inventory: The skill documents capabilities for database writes (ctx.store.upsert), network operations via RPC endpoints, and process execution through the sqd CLI.
Sanitization: While data is decoded using ABIs, the documentation does not explicitly detail sanitization steps for scenarios where indexed data is subsequently processed by an LLM, representing a standard injection surface for data processing skills.

subsquid