AgentSkillsCN

code-smells-and-antipatterns

在当前代码差异中检测新出现或恶化的代码异味与设计反模式，阐明其潜在危害，并提出最精简的修复方案。适用于非 trivial 的结构性变更、重构操作，或当代码可维护性出现下滑时使用。

SKILL.md
--- frontmatter
name: code-smells-and-antipatterns
description: "Detect new or worsened code smells / design anti-patterns in the current diff, explain why they matter, and propose the smallest fix. Use for non-trivial structural changes, refactors, or when maintainability regresses."
metadata:
  short-description: Smells & anti-patterns triage

Purpose

This skill is a triage tool to reduce maintainability risk by finding “reader-stoppers” at the design level, not to enable big refactors.

When to use (trigger conditions)

Agents MUST use this skill when any of the following applies:

  • •New module/subsystem or new “main” class is introduced.
  • •A change introduces or edits public APIs or cross-module boundaries.
  • •A change increases indirection/flags/config branching, or creates new “manager/service” hubs.
  • •A refactor moves logic across files or layers (controller/usecase/domain/adapters/etc.).
  • •Review request explicitly mentions: smell / anti-pattern / god object / big ball of mud / anemic domain model / shotgun surgery / etc.

If none applies, do not force it.

How to use (procedure; must be stepwise)

  1. •Identify the units impacted by the diff (files + key functions/classes).
  2. •Scan for new or worsened smells only.
  3. •Produce up to 3 findings, each with:
    • •name (smell/anti-pattern label)
    • •“why this looks like it” (evidence from diff)
    • •risk if left as-is
    • •smallest fix (prefer minimal diff)
    • •which existing skill to invoke for the fix ($code-readability, $modularity, $architecture-boundaries, $working-with-legacy-code, $error-handling, $observability, $test-driven-development)
  4. •If you choose NOT to fix now, you MUST state:
    • •why it is not new/worsened, or
    • •why fixing now would increase risk, and
    • •what to monitor (tests/metrics/logs) until a planned follow-up.

Output expectation (strict format)

Require the output to include:

Smells & Anti-patterns Review

  • •Scope: (changed units)
  • •Findings: (0–3 items)

For each finding:

Finding N: <label> (type: smell|design anti-pattern|architecture anti-pattern; impact: blocker|important|nice-to-have)

  • •Introduced/Worsened by this diff?: yes|no
  • •Evidence (diff-level):
  • •Why it matters here:
  • •Smallest fix:
  • •Skill to apply:
  • •If not fixing now (only allowed if not introduced/worsened): justification + follow-up note

If there are 0 findings:

  • •State: “No new/worsened smells found” and list what you checked.