Writing Plans

Overview

Write comprehensive implementation plans assuming the engineer has zero context for our codebase and questionable taste. Document everything they need to know: which files to touch for each task, code, testing, docs they might need to check, how to test it. Give them the whole plan as bite-sized tasks. DRY. YAGNI. TDD. Frequent commits.

Assume they are a skilled developer, but know almost nothing about our toolset or problem domain. Assume they don't know good test design very well.

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

Context: This should be run in a dedicated worktree (created by brainstorming skill).

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

Bite-Sized Task Granularity

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

•"Write the failing test" - step
•"Run it to make sure it fails" - step
•"Implement the minimal code to make the test pass" - step
•"Run the tests and make sure they pass" - step
•"Commit" - step

Plan Document Header

Every plan MUST start with this header:

markdown

# [Feature Name] Implementation Plan

> **For Claude:** Spawn `task-builder` agent to implement this plan task-by-task.

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

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

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

---

Task Structure

markdown

### Task N: [Component Name]

**Parallel:** yes | no
**Blocked by:** [Task IDs or none]
**Owned files:** `path/to/file.ts`, `path/to/test.ts`

**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"

If Parallel: yes (task-builder):

•Replace commit step with: "Return diff summary for orchestrator review (no commit)"

code


## Remember
- Exact file paths always
- Complete code in plan (not "add validation")
- Exact commands with expected output
- Reference relevant skills with @ syntax
- Include **Parallel**, **Blocked by**, and **Owned files** for every task
- Validate **Owned files** do not overlap across parallel tasks (use a quick table or script to check for duplicates)
- DRY, YAGNI, TDD, frequent commits

## Owned Files Validation

Use a quick duplicate check before running parallel tasks (portable across macOS/Linux):

```bash
rg '\*\*Owned files:\*\*' IMPLEMENTATION_PLAN.md \
  | sed 's/.*\*\*Owned files:\*\* *//' \
  | tr ',' '\n' \
  | sed 's/`//g' \
  | sed 's/^[[:space:]]*//;s/[[:space:]]*$//' \
  | grep -v '^$' \
  | sort \
  | uniq -d

If this outputs any paths, fix overlaps before parallelizing.

Example Parallel Task

markdown

### Task 3.2: Add User Authentication

**Parallel:** yes
**Blocked by:** Task 3.1
**Owned files:** `src/auth/auth.ts`, `src/auth/auth.test.ts`, `src/middleware/auth-middleware.ts`

**Files:**
- Create: `src/auth/auth.ts`
- Modify: `src/middleware/auth-middleware.ts:12-48`
- Test: `src/auth/auth.test.ts`

Execution Handoff

After saving the plan, offer execution choice:

"Plan complete and saved to docs/plans/<filename>.md. Two execution options:

1. Execute Now (this session) - I spawn task-builder agent per task, review between tasks, fast iteration

2. Parallel Session (separate) - Open new session in worktree, spawn task-builder agent, batch execution with checkpoints

3. Parallel Tickets - Mark tasks with Parallel/Blocked by/Owned files, create worktrees, run /task-builder per task, review diffs before merge

Which approach?"

If Execute Now chosen:

•Spawn task-builder agent with the plan path
•Stay in this session
•Fresh agent per task + code review

If Parallel Session chosen:

•Guide them to open new session in worktree
•Spawn task-builder agent with the plan path

If Parallel Tickets chosen:

•Validate plan tasks include Parallel/Blocked by/Owned files
•Run the Owned Files Validation script to check for overlaps
•Create one worktree per parallel task
•Run /task-builder for each worktree
•Review diffs + tests in worktrees before merging
•Prefer task-builder when tasks share files or require shared context