making-plans
Use when design is complete and you need detailed implementation tasks - breaks epics into coarse-grained Beans issues with TDD guidance, exact file paths, and verification steps
When & Why to Use This Skill
The 'making-plans' Claude skill bridges the gap between high-level software design and technical execution. It automates the process of breaking down complex epics into granular, actionable tasks (Beans issues), providing developers with precise Test-Driven Development (TDD) guidance, exact file paths, and standardized verification steps to ensure high-quality code implementation and streamlined project tracking.
Use Cases
- Converting a finalized architectural design into a structured backlog of bite-sized implementation tasks for an engineering team.
- Onboarding developers to a complex codebase by providing explicit step-by-step instructions, including failing test cases and specific file modifications.
- Enforcing Test-Driven Development (TDD) and consistent commit standards across a project by generating detailed task descriptions with built-in verification checklists.
- Synchronizing project management tools by automatically creating child issues under specific Linear epics with full technical context.
| name | making-plans |
|---|---|
| description | Use when design is complete and you need detailed implementation tasks - breaks epics into coarse-grained Beans issues with TDD guidance, exact file paths, and verification steps |
Making 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, code examples, testing approach, verification steps, docs they might need to check. But instead of a single big plan document, split the entire plan into bite-sized tasks, one Beans issue ("bean") per task. Each bean should represent a logical unit of work (e.g., "Implement auth middleware with TDD") with detailed guidance in the description.
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 making-plans skill to create implementation tasks."
⚠️ PREREQUISITE: Linear ↔ Beans Epic
BEFORE continuing, you MUST read your issue-tracking-with-beans-and-linear skill.
Before breaking down tasks, you MUST have:
- A Linear ticket (e.g., ZCO-123)
- A Beans epic referencing that ticket in its title, e.g. "ZCO-123 - "</li> </ol> <p>If either is missing, STOP and create them first. Use <code>brainstorming</code> skill if no design exists yet.</p> <h2>Task Granularity</h2> <p><strong>Each Beans issue is one logical unit:</strong></p> <ul> <li>"Implement auth middleware with TDD" — issue</li> <li>"Add user model and migrations" — issue</li> <li>"Create login endpoint with validation" — issue</li> </ul> <p><strong>Within each issue description, include step-by-step guidance:</strong></p> <ol> <li>Write the failing test</li> <li>Run it to verify failure</li> <li>Implement minimal code to pass</li> <li>Run tests to verify</li> <li>Commit</li> </ol> <h2>Beans Issue Structure</h2> <p>For each task, create a Beans issue:</p> <pre><code class="language-bash">beans create "Implement <component>" --type task --parent <epic-id> --body "<description>" </code></pre> <p><strong>Issue description should include:</strong></p> <pre><code class="language-markdown">**Files:** - Create: `exact/path/to/file.py` - Modify: `exact/path/to/existing.py:123-145` - Test: `tests/exact/path/to/test.py` **Steps:** 1. Write failing test: ```python def test_specific_behavior(): result = function(input) assert result == expected ``` Run: `pytest tests/path/test.py::test_name -v` Expected: FAIL 2. Implement: ```python def function(input): return expected ``` Run: `pytest tests/path/test.py::test_name -v` Expected: PASS 3. Commit: ```bash git commit -m "feat: add specific feature" -- tests/path/test.py src/path/file.py ``` **Verification:** - [ ] Tests pass - [ ] No type errors - [ ] Committed </code></pre> <h2>Remember</h2> <ul> <li>Exact file paths always</li> <li>Complete code examples (not "add validation")</li> <li>Exact commands with expected output</li> <li>Each task issue is a child of the epic</li> </ul> <h2>Execution Handoff</h2> <p>After creating all issues:</p> <pre><code class="language-bash">beans list --links parent:<epic-id> </code></pre> <p><strong>"Tasks created under epic <code><epic-id></code>. Run <code>beans list --status open</code> to see unblocked work. Want me to start implementing?"</strong></p> <p>When implementing:</p> <ul> <li><code>beans update <id> --status in-progress</code> before starting</li> <li><code>beans update <id> --status done</code> when done</li> </ul>