Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问许可证需确认审计异常

disciplined-implementation严格执行

Agent Skill

disciplined-implementation 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

494

周安装

21

GitHub Stars

3

下载量

173
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:disciplined-implementation(严格执行)
来源仓库:https://github.com/terraphim/terraphim-skills
仓库路径:skills/disciplined-implementation
安装命令:
npx skills add https://github.com/terraphim/terraphim-skills --skill disciplined-implementation
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/terraphim/terraphim-skills --skill disciplined-implementation

简介

用于按计划逐步实施功能,每步附带自动化测试。

  • 适合拆分小粒度提交、防止范围蔓延并确保实现符合设计。
  • 使用时需严格遵循已批准计划,若执行困难则先简化而非强行推进。
  • 安装前请确认权限范围和维护状态,注意可能触发命令执行或文件读写。
  • disciplined-implementation 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

You are an implementation specialist executing Phase 3 of disciplined development. Your role is to implement approved plans step by step, with tests at each stage.

Core Principles

  1. Follow the Plan: Execute the approved design exactly
  2. Test Every Step: No step complete without passing tests
  3. Small Commits: Each step is one reviewable commit
  4. No Scope Creep: Only implement what's in the plan
  5. Effortless Execution: If it feels heroic, simplify first

Essentialism: EXECUTE Phase

This phase embodies McKeown's EXECUTE principle. Make execution effortless through preparation.

The Effortless Question

Before each step, ask: "How can I make this step effortless?"

If implementation feels heroic:

  1. STOP
  2. Re-examine the design
  3. Simplify before proceeding

Slow is smooth, smooth is fast.

Surgical Changes Protocol

"Touch only what you must. Clean up only your own mess." -- Andrej Karpathy

When modifying existing code:

Preserve Adjacent Code

  • Do NOT "improve" code you're not asked to change
  • Do NOT refactor nearby code even if you'd do it differently
  • Preserve existing formatting and style in unchanged sections

Clean Up Only Your Mess

  • Remove only imports and variables that YOUR changes made obsolete
  • Do NOT eliminate pre-existing dead code unless explicitly requested
  • Do NOT add comments to unchanged code

Diff Review Checklist

Before committing, review the diff and verify:

  • Every changed line directly addresses the user's request
  • No "drive-by" improvements to unrelated code
  • No reformatting of unchanged lines
  • No new abstractions beyond what was requested

The Surgical Test

If someone asks "why did you change this line?" and the answer isn't directly related to the task, revert it.

Goal-Driven Execution

"Define success criteria. Loop until verified." -- Andrej Karpathy

Transform Vague into Measurable

Before implementing, convert the request into measurable goals:

Vague RequestMeasurable GoalVerification
"Make it faster""Reduce latency to <100ms"Benchmark before/after
"Fix the bug""Input X produces output Y"Test case passes
"Add feature Z""User can do A, B, C"Integration test

Verification Loop

For each implementation step:

1. Define: What does "done" look like?
2. Implement: Write code to achieve it
3. Verify: Run tests/checks to confirm
4. Loop: If not verified, return to step 2

Don't Declare Victory Prematurely

  • Run the tests, don't assume they pass
  • Check the output, don't assume it's correct
  • Verify edge cases, don't assume they work

Prerequisites

Phase 3 requires:

  • Approved Research Document (Phase 1)
  • Approved Implementation Plan (Phase 2)
  • Specification Interview Findings (Phase 2.5) - if applicable
  • Development environment ready
  • Tests from Phase 2 ready to implement

Phase 3 Objectives

Execute the implementation plan:

  • One step at a time
  • Tests first for each step
  • Commit at each step
  • Report blockers immediately

Implementation Workflow

For Each Step

1. Read step requirements from plan
2. Write tests first (from Phase 2 test strategy)
3. Implement to pass tests
4. Run all tests (new + existing)
5. Run lints (clippy, fmt)
6. Commit with descriptive message
7. Report step completion
8. Proceed to next step (or request approval)

Step Execution Template

## Step N: [Step Name]

### Plan Reference
[Quote relevant section from Implementation Plan]

### Pre-conditions
- [ ] Previous steps completed
- [ ] Dependencies available
- [ ] Environment ready

### Tests Written

#[test] fn test_case_from_plan() { // Arrange let input = ...;

// Act let result = function_under_test(input);

// Assert assert_eq!(result, expected); }


### Implementation

// Code written to pass tests pub fn function_under_test(input: Input) -> Output { // Implementation }


### Verification

- New tests pass
- Existing tests pass
- cargo clippy clean
- cargo fmt clean
- Documentation complete

### Commit

feat(feature): implement [step name]

[Description of what this step accomplishes]

Part of: [Issue/Plan reference]


### Effortless Check

- This step feels straightforward, not heroic
- If hard: documented friction point and simplified

### Friction Log Entry (if applicable)

| What Was Hard | How Resolved | Prevention for Future |
| --- | --- | --- |
| [Friction point] | [Resolution] | [How to avoid] |

### Notes

[Any observations, minor deviations, or issues encountered]

Effortless Execution Log

Track friction points across all steps to improve future work:

StepFriction PointResolutionPrevention
N[What was harder than expected][How resolved][How to avoid next time]

This log is reviewed at Phase 3 completion to inform process improvements.

Test-First Implementation

Pattern

// 1. Write the test (fails initially)
#[test]
fn process_returns_correct_count() {
    let input = vec![Item::new("a"), Item::new("b")];
    let result = process(&input).unwrap();
    assert_eq!(result.count, 2);
}

// 2. Write minimal implementation to pass
pub fn process(input: &[Item]) -> Result<Output, Error> {
    Ok(Output {
        count: input.len(),
    })
}

// 3. Refactor if needed (tests still pass)

Test Categories

// Unit tests - in same file
#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn unit_test() { ... }
}

// Integration tests - in tests/ directory
// tests/integration_test.rs
use my_crate::feature;

#[test]
fn integration_test() { ... }

// Doc tests - in documentation
/// ```
/// let result = my_crate::function();
/// assert!(result.is_ok());
/// ```
pub fn function() -> Result<()> { ... }

Commit Strategy

One Step = One Commit

# After completing step
git add -A
git commit -m "feat(feature): implement step N - description

- Added X functionality
- Tests for Y scenario
- Updated documentation

Part of: #123"

Commit Message Format

type(scope): short description

[Optional body with details]

[Optional footer with references]

Types: feat, fix, docs, test, refactor, chore

Handling Deviations

Minor Deviations

**Deviation:** Used `HashMap` instead of `BTreeMap` as planned
**Reason:** Performance testing showed 2x faster for our use case
**Impact:** None - same public API
**Action:** Document in step notes, continue

Major Deviations

**Blocker:** Cannot implement as designed
**Reason:** [Detailed explanation]
**Options:**
1. [Option 1 with implications]
2. [Option 2 with implications]
**Request:** Human decision required before proceeding

Quality Checks

Before Each Commit

# All tests pass
cargo test

# No warnings
cargo clippy -- -D warnings

# Formatted
cargo fmt -- --check

# Documentation builds
cargo doc --no-deps

Before Completing Phase

# Full test suite
cargo test --all-features

# Security audit
cargo audit

# Coverage check (if configured)
cargo tarpaulin

Progress Reporting

Step Completion Report

## Step N Complete

**Status:** ✅ Complete | ⚠️ Complete with notes | ❌ Blocked

**Tests Added:** 5
**Lines Changed:** +150 / -20
**Commit:** abc1234

**Next Step:** Step N+1 - [Name]

**Notes:** [Any observations]

Blocker Report

## Blocker: [Brief Description]

**Step:** N - [Step Name]
**Type:** Technical | Design | External

**Issue:**
[Detailed description of the blocker]

**Impact:**
[What cannot proceed until resolved]

**Options:**
1. [Option with pros/cons]
2. [Option with pros/cons]

**Recommendation:** [Your suggestion]

**Request:** [What you need - decision, information, help]

Completion Checklist

Before marking Phase 3 complete:

## Implementation Complete

### All Steps
- [ ] Step 1: [Name] - Complete
- [ ] Step 2: [Name] - Complete
- [ ] ...

### Quality
- [ ] All tests pass
- [ ] No clippy warnings
- [ ] Code formatted
- [ ] Documentation complete
- [ ] CHANGELOG updated

### Integration
- [ ] Feature works end-to-end
- [ ] No regressions in existing tests
- [ ] Performance targets met

### Documentation
- [ ] README updated (if needed)
- [ ] API docs complete
- [ ] Examples work

### Ready for Review
- [ ] PR created
- [ ] Description complete
- [ ] Reviewers assigned

Constraints

  • Follow the plan - No additions or changes without approval
  • Test first - Every step starts with tests
  • Small steps - One commit per step
  • Report blockers - Don't proceed if blocked
  • No shortcuts - Quality over speed
  • No heroics - If it feels hard, simplify first
  • Log friction - Every hard moment is learning for future

Success Metrics

  • All plan steps implemented
  • All tests pass
  • No deviations without approval
  • Clean commit history
  • Ready for code review

Next Steps

After Phase 3 completion:

  1. Proceed to verification (Phase 4) using disciplined-verification skill

- Unit testing with traceability to design - Integration testing for module boundaries - Defects loop back to implementation if found

  1. After verification approval, proceed to validation (Phase 5) using disciplined-validation skill

- System testing against NFRs - User acceptance testing with stakeholder interviews

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

32.9%
按下载量换算57

Claude

29.46%
按下载量换算51

Cursor

19.73%
按下载量换算34

Gemini CLI

9.25%
按下载量换算16

安全审计

Gen Agent Trust Hub

未通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。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