The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process external data sources which could contain malicious instructions.
Ingestion points: The skill ingests untrusted client-sent metadata (referenced in implementation-playbook.md) and processes Protobuf schema files provided by the user (described in SKILL.md).
Boundary markers: The instructions lack specific prompt delimiters for external schema content, though they do advise allowlisting metadata keys.
Capability inventory: The skill uses command-line tools for code generation (buf generate, buf lint) and performs network operations via gRPC server/client implementations.
Sanitization: The skill explicitly recommends validating and allowlisting metadata keys used for authorization or routing.
[COMMAND_EXECUTION]: The skill involves the execution of local development tools.
The guide instructs running buf lint and buf generate to validate schemas and produce Go code. These are standard industry tools for gRPC development.
[EXTERNAL_DOWNLOADS]: The configuration examples reference remote resources from a well-known service.
The buf.gen.yaml example fetches Protobuf plugins from the official buf.build registry during the code generation process.

grpc-golang