AgentSkillsCN

validate

并行运行代码校验（lint、类型检查、测试、构建）。当开发者希望在提交代码前或完成修改后进行代码校验时使用。

SKILL.md
--- frontmatter
name: validate
description: Run code validation (lint, type-check, test, build) in parallel. Use when developer wants to validate code before committing or after making changes.
user-invocable: true

/validate

Run all validation commands in parallel and report results.

When to Run

Run /validate at the end, right before committing:

  • •After ALL implementation is complete
  • •After ALL tests pass individually
  • •As final verification before creating a commit/PR

During development, use targeted commands instead:

bash
npx vitest run src/path/to/file.test.tsx   # Run specific test
npx biome check --write ./src              # Lint + auto-fix
npx tsc --noEmit                           # Type check
npx rspeedy build                          # Build JS bundle

Usage

code
/validate

Instructions

Step 1: Run all checks in parallel

Spawn 4 runner agents in parallel using the Task tool with subagent_type: "runner". All 4 MUST be spawned as parallel Task tool calls in a single message.

CheckCommand
Lintnpx biome check --write ./src && npx biome check ./src
Typesnpx tsc --noEmit
Testsnpx vitest run
Buildnpx rspeedy build

Example Task call:

code
Task tool:
  subagent_type: "runner"
  description: "Run lint check"
  prompt: "Run command: npx biome check --write ./src && npx biome check ./src"

Step 2: Report results

After all agents complete, present a summary table:

code
Validation Results:
| Check       | Status |
|-------------|--------|
| Lint        | PASS / FAIL |
| Types       | PASS / FAIL |
| Tests       | PASS / FAIL |
| Build       | PASS / FAIL |
  • •If ALL pass: report success
  • •If ANY fail: show the error summary from the failed agent(s) and suggest fixes