lets-plan-code

Files (2)

• SKILL.md
• openai.yaml · agents/

SKILL.md (first lines)

# Let's plan code

Run a planning session, using the `domain-modeling` skill.

## Wayfinder use

When `wayfinder` calls this skill, apply the active effort's exploration boundary.

Write only under `.scratch/wayfinder/<effort>/`. Do not modify `CONTEXT.md`, ADRs, source files, issue trackers, branches, worktrees, or commits.

Record a required glossary update or ADR as an implementation follow-up in the active Wayfinder ticket and map.

Return the plan to Wayfinder for storage. Do not create a separate project plan.

<what-to-do>
1. Interview the user relentlessly about every aspect of this plan.
2. Ask one concise question at a time. Use the host's user-input tool when it is available.
3. Order questions by leverage: resolve the decision that most constrains downstream choices first, then re-evaluate.
4. If a question can be answered by exploring the codebase, explore the codebase instead.
5. Stop when no unresolved decisions would change the implementation, or remaining questions are cosmetic.
6. Write a concise plan in bullets. Skip filler.
7. When the plan is complete you MUST run `fresh-eyes` (a review pass that helps catch bad assumptions). Integrate its findings into your plan. Return to step 2 if it raises new questions.
8. Show a summary of how many decisions I made this session.
</what-to-do>

## Supporting information

- If `docs/agents/issue-tracker.md` is missing from this repository, run `setup` first.
- Use the `domain-modeling` skill to plan the domain model.

From terebentina/skills, MIT.

More from terebentina/skills