Phx Plan

Plan Elixir/Phoenix Feature

Plan a feature by researching the relevant Elixir/Phoenix concerns, then output a structured plan with checkboxes.

What Makes $elixir-phoenix:phx-plan Different from /plan

1. Covers relevant concerns through resumable research tracks
2. Plans with [ecto], [liveview], [oban] task routing
3. Checks for Iron Law compliance in the plan
4. Includes mix compile/format/credo/test verification
5. Understands Phoenix context boundaries

Usage

$elixir-phoenix:phx-plan Add user avatars with S3 upload
$elixir-phoenix:phx-plan .claude/plans/notifications/reviews/notifications-review.md
$elixir-phoenix:phx-plan Implement notifications --depth deep
$elixir-phoenix:phx-plan .claude/plans/auth/plan.md --existing

Arguments

• Text after the skill name = feature description, review file, or existing plan
• --depth quick|standard|deep = Planning depth (auto-detected)
• --existing = Enhance an existing plan with deeper research

Workflow

1. Gather context — File path (skip to research), brainstorm interview.md (skip clarification), clear description, or vague
2. Clarify if vague — Ask questions ONE at a time (skip if brainstorm interview.md exists with Status: COMPLETE)
3. Detect depth — Auto-detect quick/standard/deep
4. Create research state — Before research, create .claude/plans/{slug}/scratchpad.md with a concern-track checklist
5. Gather optional runtime context — Only when Tidewave tools are independently configured and exposed; otherwise inspect source, routes, schemas, and tests
6. Research selectively — Cover only relevant concerns. Native generic subagents may run independent tracks in parallel, but they are optional. Without them, run every selected track sequentially in this session and save evidence under .claude/plans/{slug}/research/
7. Finish ALL research tracks — Maintain the scratchpad checklist, marking each selected track [x] only after its evidence is captured. NEVER write the plan while any selected track remains unchecked
8. Breadboard (LiveView) — Produce the system map from collected evidence
9. Completeness check — required when planning from review
10. Split decision — One plan or multiple, concrete options
11. Generate plan — Checkboxes, phased tasks, code patterns. Reuse .claude/plans/{slug}/scratchpad.md for decisions and dead-ends
12. Self-check (deep only) — Three questions in Risks section
13. Present and ask — STOP, show summary, let user decide

When planning from review: Every finding must appear in the plan — either as a task OR explicitly deferred by the user.

See references/planning-workflow.md for detailed step-by-step.

--existing Mode (Deepening)

Enhance an existing plan without relying on named agents:

1. Load the plan and create or update .claude/plans/{slug}/scratchpad.md
2. Identify thin sections and add a checklist of relevant concern tracks
3. Complete every track in this session, sequentially by default; generic workers are optional only for independent tracks and must write evidence under .claude/plans/{slug}/research/
4. Produce breadboarding and infrastructure notes directly from the gathered evidence, independent of whether workers were used
5. Add implementation detail, resolve spikes, and strengthen verification
6. Present a diff summary; never delete or silently rewrite existing tasks