AgentSkillsCN

Quality Check

根据代码变更，自动运行质量检测代理。

SKILL.md
--- frontmatter
description: Run automated quality agents based on code changes
argument-hint: [branch-name]
allowed-tools: Bash, Task, Read, Grep, Glob
model: opus
disable-model-invocation: true

Perform automated quality assurance by analyzing code changes and invoking appropriate specialized agents.

Usage:

  • •/quality-check - Compare current branch vs master (DEFAULT)
  • •/quality-check <branch-name> - Compare current branch vs specified branch

Examples:

bash
# Before creating a PR to master (most common)
/quality-check

# Compare to custom branch
/quality-check develop

Process:

  1. •

    Identify Changed Files

    • •Determine target branch: argument provided or "master" (default)
    • •Get changed files: git diff --name-status $TARGET_BRANCH...HEAD
    • •Note: Uses three-dot syntax (...) to compare from common ancestor (uses local branch state)
  2. •

    Apply Decision Tree (determine which agents to invoke)

    Skip ALL agents when changes are ONLY:

    • •Config/settings changes (no logic)
    • •Typo/comment/PHPDoc fixes
    • •Pure renames (no behavior change)
    • •Translation updates (POT files)
    • •Asset changes (CSS/images without logic changes)
    • •NOTE: JS/JSX files in template/js/ are React source code, NOT asset changes — always evaluate them

    Invoke agents based on these criteria:

    quality-security - when touching:

    • •Authentication/authorization code (capabilities, user permissions)
    • •Database queries (ANY query modifications, especially wpdb operations)
    • •User input handling (sanitization, validation)
    • •File uploads/downloads
    • •AJAX endpoints and REST API endpoints
    • •Nonce generation/verification
    • •WordPress hooks that handle user data
    • •Media file operations

    quality-code-efficiency - when:

    • •Adding new business logic (PHP or JS/React)
    • •Modifying existing methods with logic
    • •Writing verbose code that could be more concise
    • •NOTE: Detects both duplication AND verbosity (framework-agnostic patterns)

    quality-performance - when touching:

    • •Database queries (ANY modifications, wpdb operations)
    • •Loops over collections (especially media file scanning)
    • •File I/O operations
    • •External API calls
    • •Caching logic
    • •React component rendering (unnecessary re-renders)

    quality-patterns - when touching:

    • •Files in src/ with WordPress plugin code (Controllers, Models, Services, etc.)
    • •Bootstrap/initialization files
    • •WordPress hooks and filters implementation
    • •REST API route registration and handlers
    • •React components using WordPress packages
    • •NOTE: Skip config-only changes or simple hook registrations without logic

    quality-architecture - when:

    • •Creating new classes/services
    • •Modifying 2+ related files
    • •Cross-namespace changes (e.g., Core + Admin + Api)
    • •Refactoring that changes structure

    quality-test-coverage - when:

    • •Adding new business logic (methods with logic)
    • •Modifying existing business logic behavior
    • •Adding new AJAX/REST endpoints
    • •Adding database operations (wpdb queries, entity persistence)
    • •Fixing bugs (needs regression test)
    • •NOTE: Skip config-only changes, typos, comments, simple getters/setters

    quality-test-rigor - when:

    • •Adding new tests (PHP or JS)
    • •Modifying existing tests
    • •NOTE: Validates test quality and rigor, not coverage

    quality-maintainability - ALWAYS

    • •Detects technical debt markers, magic values, complexity issues, and code hygiene problems

    quality-comments - ALWAYS

    • •Validates that code comments are meaningful and not redundant

    quality-docs - ALWAYS

    • •Analyzes code changes to determine if documentation updates are needed
  3. •

    Invoke Agents in Parallel

    • •Use Task tool to launch all matching agents in a SINGLE message
    • •Each agent gets the list of modified files for context
    • •Agents run independently and in parallel
    • •Multiple agents may analyze the same file (e.g., security + performance on queries) - this is intentional
  4. •

    Present Consolidated Report

    • •Analyze and consolidate all agent outputs
    • •Summary of which agents ran
    • •Key findings from each agent (prioritized by severity)
    • •Cross-reference findings between agents (e.g., security + performance on same code)
    • •Actionable recommendations
    • •Overall quality score

Decision Rule:

  • •Default to YES: When uncertain if an agent should run, invoke it (prefer false positives over false negatives)
  • •Run in Parallel: Always invoke all matching agents in a single message for efficiency
  • •Avoid over-engineering: Prefer simple, proportional fixes over complex abstractions

Important:

  • •This skill should be run AFTER completing code changes, not during
  • •Agents analyze code quality, they don't modify code
  • •Results are advisory - you decide which recommendations to implement