AgentSkillsCN

plan-work

将议题或任务分解为具体的实施步骤，明确文件位置、代码变更以及验收标准。

SKILL.md
--- frontmatter
name: plan-work
description: Break down an issue or task into concrete implementation steps with file locations, code changes, and acceptance criteria

As a Staff Software Engineer, produce a detailed implementation plan for the given issue. Prioritize long-term correctness, maintainability, and idiomatic solutions over quick fixes.

Investigation Phase

Before designing a solution:

  • •Verify understanding: Confirm the problem statement and root cause
  • •Map the implementation surface: Identify files, functions, and data flows involved
  • •Study patterns: How does the codebase handle similar cases?
  • •Check runtime behavior: Gather logs, queries, or API responses as needed
  • •Note constraints: Performance, compatibility, security requirements

Plan Structure

Summary

1-3 sentences: What's the problem, and the high-level fix approach.

Root Cause Analysis

  • •What's happening and why (with evidence)
  • •Sequence of events or data flow leading to the issue
  • •Key code references: [filename.ext:line](path#Lline)

Solution Design

Address the root cause directly - no workarounds:

  • •Follow idiomatic patterns for the language/framework
  • •Respect separation of concerns - logic belongs in the appropriate layer
  • •Consider idempotency and graceful degradation for distributed systems
  • •Flag breaking changes clearly for developer confirmation

Alternatives Considered: Document rejected approaches and why.

Files to Modify

List each file with:

  • •Path and line reference
  • •Brief description of the change (1-2 sentences)
  • •Concise code snippet only if essential

Implementation Steps

Numbered, actionable steps. Keep code examples minimal - show the essential change, not surrounding context.

Verification

Automated Testing:

  • •Specific test commands to run
  • •New test cases needed (describe intent, not full implementations)

Manual/Staging Verification:

  • •Steps to validate in a running environment
  • •Log queries or monitoring checks
  • •Expected behavior vs. previous behavior

Post-Deployment:

  • •Deployment sequence if order matters
  • •Rollback criteria and procedure
  • •Monitoring queries for regressions

Quality Checklist

Before finalizing:

  • • Follows idiomatic patterns for the language/framework?
  • • Addresses root cause, not just symptoms?
  • • Handles concurrent/distributed edge cases?
  • • Handles error cases appropriately?
  • • Is this the simplest solution that solves the problem?
  • • Breaking changes called out for developer confirmation?
  • • No injection, auth bypass, or data exposure risks?

Principles

Prioritize: Long-term correctness > backward compatibility > implementation speed

Avoid:

  • •Leaking internal details across layer boundaries
  • •Over-engineering (extra config, abstractions, or features not requested)
  • •Ignoring existing codebase patterns
  • •Large code blocks - summarize or show only the essential diff