# Elsa Work Unit Planner

> Turn Elsa-brain findings into an architecture-first or Speckit-ready work-unit plan. Use when a report finding, deferred decision, weak implementation, docs/maps/skills gap, or user-selected next unit needs scope, success criteria, source-of-truth layer, and exact files before implementation.

- Skill: `elsa-workflows/elsa-work-unit-planner` (Agent Skill)
- Install (CLI): `npx skillmds@latest add elsa-workflows/elsa-work-unit-planner`
- Raw SKILL.md: https://api.skillmd.com/api/skills/elsa-workflows/elsa-work-unit-planner/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: elsa-workflows (https://skillmd.com/u/elsa-workflows)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/elsa-workflows/elsa-work-unit-planner

---


## User Input

```text
$ARGUMENTS
```

## Outline

1. Read `AGENTS.md` for source-of-truth layers and the program goals / drift guard.
2. Read `docs/skills/catalog.md#work-unit-planner` and `docs/skills/catalog.md#promote-finding-to-work-unit`.
3. Inspect `docs/program-goals/` and identify whether the work belongs in a named bucket or should remain `none/free-flow`.
4. Read the source report/spec/constitution section behind the finding.
5. Classify the unit as architecture development, feature development, codebase verification, docs/maps/skills work, or code.
6. Define goal, success criteria, in scope, out of scope, source-of-truth layer, program-goal route, program-goal state when used, affected gates/maps/docs, tests, and review points.
7. End with a `<proposed_plan>` block listing exact files to create or update, including likely follow-through obligations such as tests, extension-point catalogs, generated-map refreshes, or docs updates.

Do not run Speckit or edit code unless the user explicitly approves implementation. After a unit is approved, required follow-through obligations are part of completing that unit unless they introduce a new architecture decision.