plan

name	plan
description	Everyday, auto-invocable planning skill for routine feature/bug/refactor work. Sonnet-tier — the non-Fable default. Use for most planning needs; escalate to /plan-deep (manual-only, Fable-5) only for the rare highest-stakes plan.
model	sonnet

/plan — everyday planning

The routine planning workhorse. It emits the same self-contained HTML plan format as plan-deep, but from the lighter shared template .claude/skills/lib/plan-template-lite.html — keeping status-marker chips, a per-phase testing-strategy loop, and optional image slots, while dropping the append-only-metadata <details> ceremony and the Questionables section (those are reserved for plan-deep's highest-stakes plans; CONSTITUTION.md N2 scopes Fable-5 to that one skill only — this skill never sets model: fable and never will).

When to use this vs. plan-deep

• Use plan (this skill, auto-invocable) for: standard features, bug fixes, refactors, most of what a /pp:run-style pipeline or ordinary conversational planning needs.
• Suggest plan-deep only when a task is genuinely major-scope, irreversible, or cross-cutting enough to warrant Fable-5's extra structure/thinking budget — and even then, per N2, you must use AskUserQuestion to get explicit approval before any Fable path proceeds (see the model-routing skill). Never invoke /plan-deep yourself; only the user can.

Template & output

Build the plan from the lighter shared template .claude/skills/lib/plan-template-lite.html — copy it to the output location and fill every «placeholder». Same self-contained HTML + embedded design system as plan-deep, minus the metadata/Questionables ceremony.

Output location (run-aware): .fable/<run_id>/artifacts/<slug>.html when a run is active, else specs/<slug>.html in the repo root; sibling <slug>/ image folder either way. <slug> is a short kebab-case name from the plan title.

Keep every structural piece the lite template ships: the status legend, an Overview, one .phase block per phase (each with its Testing-strategy loop and its [f] escape hatch), and Notes. Repeat the .phase block per phase, updating the status chips live on disk.

Images are optional for lite plans. The template's <figure> carries a <!-- {{...IMAGE: hero | … }} --> slot; if you want a diagram, fill the subject and use the same manual pipeline as plan-deep (plan-images.ps1 extract → generate the PNG externally when credits allow → apply). A slot left ungenerated renders as a tidy "image pending" placeholder, so the plan is fully usable without it. Never claim an image was rendered when only a slot/prompt exists.

Preserved Invariants vs Changed Behaviors (CONSTITUTION.md N9)

Revised from Markdown output to HTML. Preserved: status vocabulary [] / [wip] / [x] / [f]; the per-phase testing-strategy loop and its [f] escape hatch; model: sonnet (never fable, N2); still lighter than plan-deep (no metadata/Questionables ceremony). Changed: output format Markdown → self-contained HTML (.claude/skills/lib/plan-template-lite.html); optional <figure> image slots + the manual plan-images pipeline; run-aware output path.

Guidance

• Keep phases small enough that each Testing Strategy loop is genuinely checkable — a phase with no verifiable exit condition is a sign it should be split.
• Update status markers live, on disk, as work proceeds — don't let them go stale relative to actual progress.
• If you notice a plan created with this skill is turning out to need append-only audit history, a Questionables toggle, or genuinely wants Fable's extra reasoning budget, say so explicitly and let the user decide whether to restart it under /plan-deep instead of retrofitting the lightweight template.