mcpField
W
writing-plans

writing-plans

github.com/izyanrajwaniAgent workflows474 installsUpdated 2w ago
Repository

About writing-plans

A skill for creating implementation plans that engineers can follow without prior codebase context. It emphasizes exact file paths, complete code, test-first steps, and small independently verifiable tasks.

Key features

  1. 01Review requirements and codebase structure
  2. 02Specify exact paths for each operation
  3. 03Include complete code and verification commands
  4. 04Break work into short, test-first steps
  5. 05Check plan completeness and hand off execution

Use cases

  1. 01Plan implementation steps for a new feature
  2. 02Turn requirements into verifiable test tasks
  3. 03Write plans for engineers new to a codebase
  4. 04Prepare plans for subagent or new-session execution

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/izyanrajwani/agent-skills-library --skill writing-plans

After installing, type/writing-plansin Claude Code to confirm it loaded.

Details

From the project README

Writing Plans

Overview

Create implementation plans for an engineer with zero codebase context.

Each plan includes:

  • Exact file paths for every operation
  • Complete code (not "add validation here")
  • Test-first approach with verification commands
  • Bite-sized steps (2-5 min each)

Principles: DRY, YAGNI, TDD, frequent commits.

Announce at start: "I'm using the writing-plans skill to create the implementation plan."

Context: Run in dedicated worktree. If none exists, use using-git-worktrees skill first.

Save plans to: docs/plans/YYYY-MM-DD-<feature-name>.md

Before Writing

  1. Read spec/requirements completely
  2. Explore project structure (view .)
  3. Identify tech stack (package.json, pyproject.toml, etc.)
  4. Note existing patterns in similar files
  5. Check docs/ for existing conventions

Bite-Sized Task Granularity

Each step is one action (2-5 minutes), independently verifiable:

  • "Write the failing test" — step
  • "Run it to confirm failure" — step
  • "Implement minimal code to pass" — step
  • "Run tests to confirm pass" — step
  • "Commit" — step

Plan Document Header

Every plan MUST start with this header:

# [Feature Name] Implementation Plan

**Goal:** [One sentence describing what this builds]

**Architecture:** [2-3 sentences about approach]

**Tech Stack:** [Key technologies/libraries]

---

Task Structure

### Task N: [Component Name]

**Files:**
- Create: `exact/path/to/file.py`
- Modify: `exact/path/to/existing.py:123-145`
- Test: `tests/exact/path/to/test.py`

**Step 1: Write the failing test**

```python
def test_specific_behavior():
    result = function(input)
    assert result == expected
```

**Step 2: Run test to verify it fails**

Run: `pytest tests/path/test.py::test_name -v`
Expected: FAIL with "function not defined"

**Step 3: Write minimal implementation**

```python
def function(input):
    return expected
```

**Step 4: Run test to verify it passes**

Run: `pytest tests/path/test.py::test_name -v`
Expected: PASS

**Step 5: Commit**

```bash
git add tests/path/test.py src/path/file.py
git commit -m "feat: add specific feature"
```

Before Handoff

Verify plan completeness:

  • Every file path exists or will be created
  • Every command can be run exactly as written
  • No TODO/placeholder text remains
  • Tests cover all acceptance criteria from spec
  • Include exact test code, not descriptions

Execution Handoff

After saving plan, present:

"Plan saved to docs/plans/<filename>.md. Choose execution mode:

  1. Subagent-Driven — same session, fresh subagent per task, fast iteration
  2. Parallel Session — new session, batched execution with checkpoints

Which approach?"

If Subagent-Driven chosen
  • Stay in this session
  • REQUIRED SUB-SKILL: subagent-driven-development
  • Fresh subagent per task + two-stage review
If Parallel Session chosen
  • Guide user to open new session in worktree
  • REQUIRED SUB-SKILL: New session uses executing-plans

Comments

Markdown supported