ConnectRPC Skill

Build, configure, and debug ConnectRPC services and clients across Go, Python, and TypeScript (React/Angular), with Buf-based protobuf generation and practical guidance for transport choice, interceptors, and cross-language RPC workflows.

Install

npx skills add owezzy/connectrpc-skill

When this skill should trigger

Use this skill when the user is:

• building a ConnectRPC service from .proto files
• generating clients or handlers with Buf
• working with connect-go, connect-es, connect-query, or connect-python
• deciding between Connect, gRPC, and gRPC-Web
• wiring ConnectRPC into React or Angular
• debugging streaming, metadata, CORS, or code generation issues

Coverage

• Go backend: handlers, interceptors, streaming, testing, CORS, reflection
• TypeScript frontend: React (connect-query + TanStack), Angular (manual DI), interceptors, error handling
• Python backend: ASGI/WSGI, interceptors, OpenTelemetry, FastAPI mount
• Buf tooling: lint, breaking changes, BSR, managed mode, polyglot code generation, CI/CD

Structure

SKILL.md              — Routing-first workflow, defaults, gotchas, troubleshooting
references/
  go.md               — Go backend patterns
  typescript.md       — TypeScript frontend (React + Angular)
  python.md           — Python backend patterns
  buf-tooling.md      — Buf CLI and proto toolchain
scripts/
  gen.sh              — portable buf generate helper for a target repo
  validate-proto.sh   — portable buf lint + breaking helper for a target repo
evals/
  evals.json          — Starter skill evaluation prompts and assertions

Scripts

The skill now ships two portable shell scripts for use against a real ConnectRPC repository:

./scripts/gen.sh --help
./scripts/validate-proto.sh --help

Examples:

./scripts/gen.sh --dir /path/to/repo
./scripts/validate-proto.sh --dir /path/to/repo --against '.git#branch=main'

They are non-interactive, idempotent, and emit JSON to stdout so an agent can call them safely.

Security note: the scripts treat target repositories and .proto files as untrusted data. They require a trusted, already-installed buf binary and do not parse shell command strings or run network-based execution fallbacks.

Evaluation

This repo includes a starter evaluation file:

evals/evals.json

Use it to compare the skill with and without activation, then extend the prompts and assertions as the skill evolves.

Author

Owen Adirah (@owezzy)

License

MIT