AgentSkillsCN

Quality Gate Detection

为 Ralph 的自动化执行检测并配置质量门命令。适用场景包括：(1) 在分析项目以确定可用的测试、Lint 及构建脚本时；(2) 根据项目实际情况生成特定于任务的验证检查清单。这项技能使 Ralph 能够根据具体项目环境灵活调整验证策略。

SKILL.md
--- frontmatter
name: Quality Gate Detection
description: >-
  Detect and configure quality gate commands for Ralph's autonomous execution.
  Use when: (1) Analyzing a project to determine available test, lint, and build scripts,
  (2) Generatiing a task-specific verification checklist based on project reality.
  This skill enables Ralph to adapt verification to the specific project context.

Quality Gate Detection

Dynamically detect project quality gates from package.json scripts and frameworks.

Prerequisites

  • •Read access to package.json in the project root

Detection Process

Step 1: Analyze Project Scripts (Hybrid Diagnosis)

Read package.json and inspect the scripts section to identify verification commands.

Priority Heuristics:

  1. •Test: Look for test, test:unit, test:ci. Preference: npm test or specific script.
  2. •Lint: Look for lint, lint:fix, eslint. Preference: npm run lint.
  3. •Typecheck: Look for typecheck, tsc, build. Preference: npm run typecheck or tsc --noEmit.
  4. •Browser/E2E: Look for test:browser, e2e, cypress, playwright. Preference: npm run test:browser.

Framework Detection:

  • •Identify test runner: jest, vitest, mocha, ava.
  • •Identify linter: eslint, tslint, biome.
  • •Identify builder/runtime: tsc, vite, next, webpack.

Step 2: Formulate Verification Commands

Construct the actual commands to be run by the agent.

  • •If a script exists (e.g., "test": "jest"), use npm run test.
  • •If no script exists but framework is detected (e.g., jest in dependencies), try standard binary npx jest.
  • •MISSING: If no scripts/frameworks are detected for a category, leave the command empty. Do NOT assume defaults like npm test exist.

Step 3: Execution (Just-in-Time)

Use the detected commands immediately to verify your work.

  • •Do not write a quality-gates.json file.
  • •Do not ask for permission to run standard checks unless they are destructive.
  • •Simply execute the commands as part of your verification loop.

Edge Cases

Ambiguous Scripts:

  • •If multiple scripts match (e.g., test and test:all), prefer the one that seems most appropriate for local verification (often test or test:unit).
  • •When in doubt, note the ambiguity in the "rationale".

Missing Scripts:

  • •If a category (e.g., lint) has no script and no detected config, omit it from the required gates. Report it as "Not Detected".

Validation

  1. •Verify package.json exists and is valid JSON.
  2. •Ensure derived commands are executable (e.g., don't suggest npm run lint if lint script is missing).

Reference Documentation

  • •Config Template: See tools/config.json for full configuration structure