write-a-plan

name	Write-a-Plan
description	Use when a research conversation has produced something ready to implement — creates implementation plans that reference their originating conversation, carry epistemic context, and assume the implementing agent has zero codebase context

*CRITICAL* Add the following steps to your Todo list using TodoWrite:

• Read the 'Guidelines'.
• Check for a convo file. Look in docs/active/<branch>/convos/ for this session's convo file.
  • If it exists: use it as the originating conversation.
  • If it doesn't exist: "We don't have a conversation summary for this session yet. Let me create one before the plan, so the plan can reference its reasoning." Run skills/update-docs/SKILL.md first, then continue.
• Create a plan that a senior engineer can follow, with full provenance. Any absolute paths in your plan MUST take into account any worktrees that may have been created
• Think about edge cases. Add them to the plan.
• Think about questions or areas that require clarity. Add them to the plan.
• Emphasize how you will test your plan.
• Present plan to user.

Guidelines

Overview

Create implementation plans for engineers who have zero context for the codebase. Document which files to touch, what to test, what docs to check. Give them bite-sized tasks. DRY. YAGNI.

Assume the implementing engineer is talented but knows almost nothing about the toolset or problem domain.

Do not add code, but include enough detail that the necessary code is obvious.

Do not write a file to disk unless explicitly asked.

Key Difference from Generic Plans

Plans come from research conversations. They are proposals based on current understanding, not settled specifications. The implementing engineer needs to know:

• What conversation/research produced this plan
• What the current confidence level is
• What might change if we learn more

Plan Document Header

Every plan MUST start with this header:

# [Feature Name] Implementation Plan

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

**Originating conversation:** [link to docs/active/branch/convos/YYYYMMDD_name.md]

**Context:** [2-3 sentences about WHY we're doing this — what research finding or hypothesis motivated it]

**Confidence:** [How settled is this direction? E.g., "High — consistent finding across 3 pilot runs" or "Exploratory — testing whether this approach works at all"]

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

**Branch:** [Which branch to implement on]

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

---

Bite-Sized Task Granularity

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

• "Write the failing test for behavior" - step
• "Write the failing test for other behavior"
• "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

Test Section

Every plan that involves code MUST have a test section. This should be written first, and should document how you plan to test the behavior.

**Testing Plan**

I will add an integration test that ensures foo behaves like blah. The
integration test will mock A/B/C. The test will then call function/cli/etc.

I will add a unit test that ensures baz behaves like qux...

You should end EVERY testing plan section by writing:

NOTE: I will write *all* tests before I add any implementation behavior.

Your tests should NOT contain tests for datastructures or types. Your tests should NOT simply test mocks. Always test actual behavior.

Exception: Pure analysis or exploration tasks (e.g., "run this classification on 50 occupations and see what happens") do not need TDD. They need a clear description of what to run, what outputs to check, and what constitutes a surprising result.

Plan Document Footer

Every plan MUST end with this footer:

**Testing Details** [Brief description of what tests are being added and how they specifically test BEHAVIOR and NOT just implementation]

**Implementation Details** [maximum 10 bullets about key details]

**What could change:** [Things that might shift if ongoing research produces different findings — helps the implementing agent know what's provisional]

**Questions** [any questions or concerns that may be relevant that need answers]

---