meow:plan-creator

Creates plans and Architecture Decision Records (ADRs). Selects workflow model, sizes scope, validates completeness, and enforces Gate 1.

What This Skill Does

When /meow:plan or /meow:cook is invoked, this skill determines the workflow model (feature, bugfix, refactor, security), selects the right plan template by scope, drafts the plan, validates it, and presents it for Gate 1 approval. Also handles ADR generation for architecture decisions.

Core Capabilities

• Institutional memory retrieval — Reads memory/lessons.md and memory/patterns.json before planning to prevent re-solving known problems
• Structured discovery — Investigates 4 areas (architecture, existing patterns, constraints, external) before drafting
• Workflow model selection — Routes to feature, bugfix, refactor, or security model
• Template selection — Auto-routes to quick, standard, or multi-phase template by scope
• Risk classification — Maps components to LOW/MEDIUM/HIGH risk, guiding tester and reviewer prioritization
• Gate 1 enforcement — Plans must be approved via AskUserQuestion before implementation
• Context Reminder — After Gate 1 approval, prints a mode-matched cook command (e.g. meow:cook --fast in fast mode) so you never have to look up the flag
• ADR generation — Creates Architecture Decision Records in docs/architecture/
• Validation — Runs validate-plan.py to ensure plan completeness

Flag Modes

Flag	Behavior
(none)	Standard mode — full discovery + Gate 1 + validation
--auto	Auto-approves minor scope questions, still enforces Gate 1
--fast	Skip discovery, quick template only, Gate 1 still runs
--hard	Extended discovery, multi-phase template enforced
--parallel	Splits plan into parallel implementation phases

Gate 1 — AskUserQuestion

After drafting, the skill presents the plan via AskUserQuestion with three options:

• Approve — proceed, skill prints context reminder + stops
• Modify — user provides feedback, plan is revised and Gate 1 re-runs
• Reject — planning stops, no plan file is written

Output — Print & Stop

After Gate 1 approval, the skill ends with a Print & Stop:

• Prints a context reminder block with the mode-matched /meow:cook [plan path] command
• Stops — Claude will not proceed automatically
• You run the printed command when ready, or run a review skill first

Usage

/meow:plan add pagination              # → auto-selects quick template
/meow:plan build auth system           # → auto-selects standard template
/meow:plan build auth system --fast    # → skip discovery, quick template
/meow:plan redesign API --parallel     # → parallel phase plan

Skill Details

Phase: 1
Used by: planner agent Plan-First Gate: Creates plan if missing. Skips with plan path arg or --fast mode.

Gotchas

• Wrong model for task type: feature-model on a bug fix skips investigation → always confirm type first
• Goal describes activity, not outcome: "Implement OAuth" vs "Users can log in with OAuth" — next agent can't judge success → rewrite until Goal answers "what does done look like?"
• Acceptance criteria that can't be verified: "code is clean" blocks Gate 2 → every criterion must reference a specific command or file check

Related

• meow:cook — Uses plan-creator as its first step
• meow:plan-eng-review — Reviews plans created by plan-creator
• meow:plan-ceo-review — CEO-level plan review