TDD Planning
You are creating implementation plans that follow a strict Test-Driven Development approach. Every implementation step follows the TDD cycle: RED → GREEN → COMMIT/ROLLBACK.
Core TDD Principles
This skill enforces a disciplined TDD workflow:
- RED: Write a failing test first (test must fail for the right reason)
- GREEN: Write minimal code to make the test pass (nothing more)
- COMMIT/ROLLBACK: If green, commit. If stuck, rollback to last green state.
No implementation code is written without a failing test first. This is non-negotiable.
Working Agreement
These instructions establish a working agreement between you and the user. The key principles are:
- AskUserQuestion is your primary communication tool - Whenever you need to ask the user anything (clarifications, design decisions, preferences, approvals), use the AskUserQuestion tool. Don't output questions as plain text - always use the structured tool so the user can respond efficiently.
- Establish preferences upfront - Ask about user preferences at the start of the workflow, not at the end when they may want to move on.
- Autonomy mode guides interaction level - The user's chosen autonomy level determines how often you check in, but AskUserQuestion remains the mechanism for all questions.
User Preferences
Before starting planning (unless autonomy is Autopilot), establish these preferences:
Commit Granularity - Use AskUserQuestion with:
| Question | Options |
|---|---|
| "How granular should commits be?" | 1. Commit after each GREEN (Recommended for strict TDD), 2. Commit after each feature/phase completes, 3. I'll handle commits manually |
Rollback Strategy - Use AskUserQuestion with:
| Question | Options |
|---|---|
| "When a test can't be made green after reasonable effort, what's the rollback strategy?" | 1. git checkout. to last commit (Recommended), 2. git stash changes for later analysis, 3. Keep failing code, ask for guidance |
File Review Preference - Check if the file-review plugin is available (look for file-review:file-review in available commands).
If file-review plugin is installed, use AskUserQuestion with:
| Question | Options |
|---|---|
| "Would you like to use file-review for inline feedback on the plan when it's ready?" | 1. Yes, open file-review when plan is ready (Recommended), 2. No, just show me the plan |
Store these preferences and act on them during implementation.
When to Use
This skill activates when:
- User invokes a TDD-planning command
- Another skill references
**REQUIRED SUB-SKILL:** Use desplega:tdd-planning - User explicitly asks for TDD-based planning
- The feature being implemented is test-critical or safety-critical
Autonomy Mode
At the start of planning, adapt your interaction level based on the autonomy mode:
| Mode | Behavior |
|---|---|
| Autopilot | Research independently, create complete TDD plan, present for final review only |
| Critical (Default) | Get buy-in at major decision points, validate test strategy |
| Verbose | Check in at each step, confirm test approach before each cycle |
The autonomy mode is passed by the invoking command. If not specified, default to Critical.
Process Steps
Step 1: Context Gathering & Test Infrastructure Analysis
- Read all mentioned files immediately and FULLY:
- Research documents, related plans, JSON/data files - IMPORTANT: Use Read tool WITHOUT limit/offset parameters - CRITICAL: Read files yourself before spawning sub-tasks
- Analyze test infrastructure:
- What testing framework is used? (Jest, pytest, Go testing, etc.) - Where do tests live? (co-located, __tests__/, tests/, etc.) - What's the test command? (npm test, make test, pytest, etc.) - Are there existing test patterns to follow?
- Spawn initial research tasks:
- Use codebase-locator agent to find test files and test utilities - Use codebase-analyzer agent to understand current test patterns - Use codebase-pattern-finder agent to find similar test implementations - Use context7 MCP for testing library insights
- Present understanding and questions (if not Autopilot): First, present your findings as text: `
Based on the research of the codebase, I understand we need to [summary]. Test Infrastructure: - Framework: [testing framework] - Test location: [where tests live] - Test command:[command]- Existing patterns: [relevant patterns found]` Then, if there are questions, use AskUserQuestion.
Step 2: Test Strategy Design
- Identify testable units:
- What are the smallest testable pieces? - What inputs/outputs can be verified? - What edge cases matter?
- Design test progression:
- Start with simplest case (happy path) - Progress to edge cases - End with error handling
- Present test strategy (if not Autopilot): Use AskUserQuestion to validate approach: Question Options "I propose starting with [simplest test case] and progressing to [more complex cases]. Does this test progression make sense?" 1. Yes, proceed, 2. Let's discuss the order, 3. Add more test cases
Step 3: Plan Structure Development
- Create TDD cycle outline: Each feature/phase is broken into TDD cycles:
## Feature: [Name] ### Cycle 1: [Simplest behavior] - RED: Test for [specific behavior] - GREEN: Implement [minimal code] - COMMIT: "[descriptive message]" ### Cycle 2: [Next behavior]... - Get feedback on structure before writing details (unless Autopilot)
Step 4: Detailed Plan Writing
Before proceeding, exit plan mode to write the plan file.
Write the plan to thoughts/<username|shared>/plans/YYYY-MM-DD-tdd-description.md.
Path selection: Use the user's name (e.g., thoughts/taras/plans/) if known from context. Fall back to thoughts/shared/plans/ when unclear.
CRITICAL: Every TDD cycle MUST include exact test code to write and expected failure message. See template for exact format.
QA Specs (optional): Features that change user-facing behavior SHOULD include an optional ### QA Spec (optional): section at the Feature level (after the feature header, before the first TDD cycle). Unlike regular plans where QA specs attach per-phase, TDD plans attach them per-feature since the RED/GREEN cycle itself serves as the per-cycle verification. See the template for the format.
Template: Read and follow the template at cc-plugin/base/skills/tdd-planning/template.md
Step 5: Review and Iterate
- Present draft plan location: `
I've created the TDD implementation plan at:thoughts/<username|shared>/plans/YYYY-MM-DD-tdd-description.mdPlease review it.` - Iterate based on feedback (if not Autopilot)
- Offer structured review:
- After iteration, offer: "Would you like me to run /review on this TDD plan for completeness and gap analysis?" - If yes, invoke the desplega:reviewing skill on the plan document
- Learning Capture: OPTIONAL SUB-SKILL: If significant insights, patterns, gotchas, or decisions emerged during this workflow, consider using
desplega:learningto capture them via/learning capture. Focus on learnings that would help someone else in a future session. - Workflow handoff: After the plan is finalized (and optionally reviewed), use AskUserQuestion with: Question Options "The TDD plan is ready. What's the next step?" 1. Implement this plan (→
/implement-plan), 2. Run a review first (→/review), 3. Done for now (park the plan) Based on the answer:
- Implement: Suggest the /implement-plan command with the plan file path - Review: Invoke the desplega:reviewing skill on the plan document - Done: Set the plan's status to ready or parked as appropriate
- Finalize the plan - DO NOT START implementation
Review Integration
If the file-review plugin is available and the user selected "Yes" during User Preferences setup:
- After creating plans, invoke
/file-review:file-review <path> - If user selected "No" or autonomy mode is Autopilot, skip this step
TDD Cycle Requirements (MANDATORY)
Every implementation step MUST follow the TDD cycle. Plans without proper TDD cycles are incomplete.
Required Structure for Each Cycle
### Cycle N: [Behavior being implemented]
#### RED Phase
**Test to write:**// Exact test code goes here
**Expected failure:**
[Expected error message or failure output]
**Verify RED:** `[test command]` should fail with the above message
#### GREEN Phase
**Implementation approach:** [Brief description of minimal code needed]
**Files to modify:**
- `path/to/file.ext`: [what to add/change]
**Verify GREEN:** `[test command]` should now pass
#### COMMIT/ROLLBACK
**If GREEN:**
- Commit message: `"[descriptive message following conventional commits]"`
- Command: `git add -A && git commit -m "[message]"`
**If STUCK (can't reach GREEN after reasonable effort):**
- Rollback: `git checkout.` (or `git stash` based on preference)
- Reassess: Consider if the test is too big, or if design needs rethinking
Validation Checklist
Before finalizing any TDD plan, verify:
- [ ] Every implementation step has RED → GREEN → COMMIT/ROLLBACK structure
- [ ] RED phase includes exact test code to write
- [ ] RED phase includes expected failure message
- [ ] GREEN phase describes minimal implementation
- [ ] Each cycle is small enough to complete in one sitting
- [ ] Cycles build on each other progressively
- [ ] Rollback strategy is clear for each cycle
Important Guidelines
- Tests First, Always: Never describe implementation without the test that drives it
- Minimal GREEN: Only write enough code to pass the test, nothing more
- Small Cycles: If a cycle seems big, break it into smaller cycles
- Progressive Complexity: Start simple, add complexity through new tests
- Clear Rollback Points: Every commit is a safe point to return to
- No Speculative Code: Don't add code "because we'll need it later"