---
paths:
  - "**/*.ts"
  - "**/*.tsx"
---
- Use strict TypeScript (`"strict": true` in tsconfig)
- Prefer `interface` over `type` for object shapes
- Use explicit return types on exported functions

---
paths:
  - "**/*.test.ts"
  - "**/*.spec.ts"
---
- Use Vitest for unit tests
- Mock external APIs, never call them in tests
- Aim for >80% coverage on business logic

Use TypeScript with strict mode. Follow ESLint rules.
Run `npm test` before suggesting changes are complete.

---
name: deploy
description: Deploy to production
allowed-tools: Bash
---
Run `npm run build && npm run deploy`.

---
name: reviewer
description: Code review agent
model: sonnet
tools: Read, Bash
---
Review all TypeScript changes for correctness, type safety, and test coverage.

Use Python 3.12+. Follow PEP 8 and type all functions.
Run `pytest` and `ruff check .` before suggesting changes.

---
paths:
  - "**/*.py"
---
- Add type annotations to all function signatures
- Prefer `list[str]` over `List[str]` (3.12+ native generics)
- Use `TypedDict` for complex dict structures

---
paths:
  - "**/test_*"
  - "**/*_test.py"
---
- Use pytest with fixtures
- Use `pytest.raises` for expected exceptions
- Mock external services with `unittest.mock`

---
paths:
  - "**/api/**"
  - "**/routes/**"
---
- Use Pydantic models for request/response validation
- Return proper HTTP status codes
- Add OpenAPI descriptions to endpoints

---
name: venv
description: Set up virtual environment
allowed-tools: Bash
---
Run `python -m venv .venv && source .venv/bin/activate && pip install -e ".[dev]"`.

---
name: security-reviewer
description: Review for security vulnerabilities in Python code
model: sonnet
tools: Read, Bash
---
Check for SQL injection, SSRF, path traversal, and insecure deserialization.
Run `bandit -r src/` and review the results.

Use idiomatic Rust. Run `cargo clippy -- -D warnings` and `cargo test`
before suggesting changes are complete.

---
paths:
  - "**/*.rs"
---
- Use `anyhow::Result` for application code, `thiserror` for libraries
- Never use `.unwrap()` in production code — use `?` or `.expect("reason")`
- Return `Result` from all public functions that can fail

---
paths:
  - "**/tests/**"
  - "**/*_test.rs"
---
- Use `#[test]` for unit tests, `tests/` directory for integration
- Use `assert_eq!` with descriptive messages
- Test error cases, not just happy paths

- Every `unsafe` block must have a `// SAFETY:` comment
- Prefer safe abstractions — only use unsafe when necessary
- Audit all `unsafe` usage before merge

---
name: release
description: Create a new release
allowed-tools: Bash
---
Run `cargo test && cargo clippy -- -D warnings`, bump version in Cargo.toml, create git tag, push.

Use idiomatic Go. Run `go test ./...` and `golangci-lint run`
before suggesting changes are complete.

---
paths:
  - "**/*.go"
---
- Always check returned errors — never use `_`
- Wrap errors with `fmt.Errorf("context: %w", err)`
- Use sentinel errors for expected cases

---
paths:
  - "**/*_test.go"
---
- Use table-driven tests
- Use `testify/assert` for assertions
- Test both success and error paths

---
paths:
  - "**/handler/**"
  - "**/api/**"
---
- Use `net/http` or chi router
- Return structured JSON errors
- Log with `slog` (structured logging)

---
name: build
description: Build and test the project
allowed-tools: Bash
---
Run `go build ./... && go test ./... && golangci-lint run`.

---
name: db-reviewer
description: Review database migrations and queries
model: sonnet
tools: Read, Bash
---
Review SQL migrations for correctness. Check for missing indexes, N+1 queries, and unsafe migrations.
Run `go test ./internal/db/...` after any migration change.