Token导航 LogoToken导航TokenDH.com
研究检索敏感数据github未标认证来源可访问许可证需确认审计通过

create-plan制定计划

Agent Skill

create-plan 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

321

周安装

13

GitHub Stars

416

下载量

101
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:create-plan(制定计划)
来源仓库:https://github.com/kasperjunge/agent-resources
仓库路径:skills/create-plan
安装命令:
npx skills add https://github.com/kasperjunge/agent-resources --skill create-plan
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/kasperjunge/agent-resources --skill create-plan

简介

用于查找和筛选相关技术资源与信息。create-plan 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

  • 适合根据关键词快速定位候选结果或研究线索。
  • 可结合任务场景匹配适用工具链和技术方案。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。
  • 使用时需提供清晰的目标描述以提高检索准确性。
  • 安装前建议确认是否需要访问外部网络资源。

SKILL.md

Create Plan

Create a comprehensive implementation plan and enter plan mode for execution.

Position in Workflow

Step 4 of development workflow:

  1. /research - Understand problem, explore implementation
  2. /brainstorm-solutions - Explore solutions
  3. /design-solution - Converge on a solution
  4. /make-plan - Create implementation plan (THIS)
  5. Code, review, ship

Core Principle

First principles. Simple over easy. Go deep.

You have a bias toward finishing quickly. Resist it.

Never cut corners or take the path of least resistance. Prefer the fundamentally right solution even if it requires significantly more work. The cost of a suboptimal solution compounds over time. Rushed plans degrade codebases.

Simple ≠ Easy:

  • Easy: Less work now, more pain later (shortcuts, hacks, "good enough")
  • Simple: More work now, less pain forever (clean, maintainable, right)

Always choose simple. Always go deep enough to find it.

Input

Default: Use chosen solution from current conversation context.

If argument provided:

  • GitHub issue number/URL: Fetch full context with scripts/gh_issue_phase.sh get-issue $ARG
  • Free-form text: Additional constraint or clarification for the plan.

Workflow

1. Verify Ready State

Before planning, confirm:

  • Research is complete (problem understood)
  • Solution space was explored
  • A solution has been chosen

If missing context, ask: "I need [research/solution choice] before creating a plan. Should I run /research or /design-solution first?"

2. Specify Behavior Explicitly

This is the most critical step. Be exhaustive.

For each feature/change, document:

Desired Behavior

  • What SHOULD happen in normal cases
  • Expected outputs for typical inputs
  • Success criteria

Undesired Behavior

  • What MUST NOT happen
  • Edge cases to handle gracefully
  • Error conditions and their responses

Usage Patterns

  • Common usage patterns (happy path)
  • Unhappy paths (errors, edge cases, invalid input)
  • Boundary conditions

Example format:

## Behavior: User Authentication

### Desired
- Valid credentials → session created, redirect to dashboard
- Session expires after 24h of inactivity
- Invalid credentials → clear error message, no session

### Undesired
- NEVER store password in plain text
- NEVER expose session token in URL
- NEVER allow infinite login attempts

### Edge Cases
- Expired password → prompt reset before login
- Concurrent sessions → configurable limit
- Network timeout during auth → retry with backoff

3. Go Deep - Gather Sufficient Context

Do not plan without full understanding. Read more code than feels necessary. Understand the implications of each decision.

Before finalizing any part of the plan:

  • Read all related files, not just the obvious ones
  • Understand how similar problems were solved elsewhere in the codebase
  • Trace the full impact of proposed changes
  • Identify hidden dependencies and side effects

You are not ready to plan if:

  • You haven't read the files your changes will affect
  • You don't understand the existing patterns in this area
  • You're making assumptions instead of verifying

4. Apply First Principles Check

For each part of the plan, challenge ruthlessly:

QuestionPurpose
Is this necessary?Avoid bloat
Is there a simpler approach?Simple > easy
Does this fight the codebase?Respect existing patterns
What would we regret in 6 months?Long-term thinking
Are we cutting corners?No shortcuts
Do I have enough context?Go deeper if uncertain

Red flags - STOP and reconsider:

  • "This is faster but..." → You're choosing easy over simple
  • "We can clean this up later..." → You won't. Do it right now.
  • "Good enough for now..." → It will become permanent tech debt
  • Adding TODO comments for "later" → Later never comes
  • Skipping error handling "for simplicity" → That's not simplicity, it's negligence
  • "I think this should work..." → You don't know. Go read more code.
  • Feeling rushed → Slow down. Bad plans cost more than slow plans.

5. Design Tests

Write test specifications BEFORE implementation details:

Test design principles:

  • Test behavior, not implementation
  • Cover happy paths AND edge cases
  • One assertion per test (when practical)
  • Tests should be fast and isolated

Format:

## Tests

### Unit Tests
- `test_valid_credentials_creates_session` - happy path
- `test_invalid_credentials_returns_error` - error case
- `test_expired_password_prompts_reset` - edge case

### Integration Tests
- `test_full_login_flow_with_redirect`
- `test_session_timeout_behavior`

6. GitHub Issue Tracking

If a GitHub issue was provided or is available from prior phases:

Post behavior specification and test design as a phase comment and set the label. This must happen BEFORE entering plan mode.

echo "$PLAN_SUMMARY" | scripts/gh_issue_phase.sh post-phase $ISSUE planning
scripts/gh_issue_phase.sh set-label $ISSUE phase:planning

Pass the issue number forward for code-review and commit phases.

7. Enter Plan Mode

After documenting behavior and tests, switch to plan mode:

Say: "Entering plan mode to create the implementation plan."

Then use the EnterPlanMode tool.

Output Format (Before Plan Mode)

## Implementation Plan: [Feature Name]

### Chosen Solution
[Brief restatement of chosen approach from /design-solution]

### Behavior Specification

#### [Component/Feature 1]

**Desired:**
- [What should happen]

**Undesired:**
- [What must not happen]

**Edge Cases:**
- [Boundary conditions and handling]

#### [Component/Feature 2]
[Same structure]

### Tests

#### Unit Tests
- `test_name` - [what it verifies]

#### Integration Tests
- `test_name` - [what it verifies]

### First Principles Check
- [Confirmation that no shortcuts are taken]
- [Confirmation of simplest viable approach]
- [Alignment with existing patterns]

---

Entering plan mode to create implementation steps.

Common Mistakes

MistakeFix
Insufficient context gatheringRead ALL related files before planning
Vague behavior specsWrite explicit examples for each case
Only happy pathAlways include unhappy paths and edge cases
Tests as afterthoughtDesign tests BEFORE implementation
Shortcut rationalizationApply first principles check ruthlessly
Rushing to plan modeComplete behavior spec first
Missing "undesired" sectionExplicitly state what must NOT happen
Planning with assumptionsVerify by reading code, don't assume
Choosing easy over simpleMore work now = less pain forever

What NOT to Do

  • Do NOT plan without reading all affected code
  • Do NOT make assumptions - verify by reading
  • Do NOT skip behavior specification
  • Do NOT leave edge cases undefined
  • Do NOT take shortcuts for "simplicity"
  • Do NOT skip the first principles check
  • Do NOT enter plan mode without tests designed
  • Do NOT proceed if solution wasn't explicitly chosen
  • Do NOT rush - bad plans cost more than slow plans
  • Do NOT rationalize shortcuts with "we can fix it later"

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

38.8%
按下载量换算39

Claude

29.66%
按下载量换算30

Cursor

18.29%
按下载量换算18

Gemini CLI

8.9%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills