Idea Pipeline (Android Indie — Feature Build Phase)

Single skill, four phases — discover (conditional) → generate 2 → refine scope → brief. Output is an idea-brief.md ready for kta-prd-pipeline. The flow is shape-the-idea, not kill-the-idea.

Scope of this skill (current phase): FEATURE BUILD ONLY. Monetization is intentionally OUT of scope here — we focus on identifying core value, picking the right feature surface to ship, and exposing risk/reward/tradeoff when scoping down. Pricing/billing/ads decisions belong to a later monetization pass after the feature value is validated.

Guiding Mantra (Surface to User in Every Phase)

Ship finished > ship perfect. Perfect never exists. Ship and improve is the key of success.

Apply YAGNI aggressively in every phase — when generating ideas, scoping MVP, picking features. Default-cut anything that doesn't serve the core value loop; defer to v2. When user adds scope, remind them: a shipped Lean MVP teaches more than a polished concept stuck in dev.

Core-value-first principle: prioritize the one feature that delivers the irreplaceable user value. Every other feature is a candidate for cut — and each cut should be reasoned through with explicit risk/reward/tradeoff (see Phase 2).

When to Use

• "I want to build an Android app but need an idea"
• "I have no idea what to build"
• "Validate this concept: …"
• "Help me find an indie dev niche"
• New feature ideation for an existing app
• Pre-PRD entry point
• "Help me scope down my idea to core value"

Step-by-Step Interaction Rule (NON-NEGOTIABLE)

This skill ALWAYS guides the user step by step via AskUserQuestion. Every choice, pick, select, flag, confirm, or rank — use the tool. Never emit raw markdown questions and wait for prompt-response. There are no modes, no levels, no toggles — the natural flow is: ask one thing at a time, the user picks an option, the skill moves forward. Pre-fill sensible defaults wherever possible so the student is selecting from options, not writing fields from scratch.

Single-select → multiSelect: false. Multi-select (e.g., flag favorites) → multiSelect: true. Always include 2–4 concrete options + a free-text "Other" option when relevant. Keep header ≤12 chars, question ≤80 chars, each option label ≤17 chars.

Applies to: trigger path, seed favorites, candidate pick, scope decisions (surface / connectivity / sharing / accounts), feature cut/keep, handoff, etc. Do NOT batch multiple decisions into one prompt — one question at a time, walk the user through.

Workflow

0. Phase 0 — Discover (conditional) — read references/phase-0-discover.md. Run ONLY if student has no idea or asks "help me find one". Pick ONE of 7 trigger paths (personal pain, hobby-driven, niche-of-niche, reskin-with-wedge, trending Play Store, daily-tool gap, sunset-tool revival) — outputs 5–10 seed concepts. WebFetch sources in references/discovery-sources.md; fallback to prompt-only on fetch failure.
1. Phase 1 — Generate — read references/phase-1-generate.md. Produce exactly 2 candidate ideas (not 3-5) with: problem, target user, core value (one sentence), MVP scope, differentiation, killer feature. No monetization fields. Use AskUserQuestion to advance one or both (or regenerate).
2. Phase 2 — Refine & Adjust — read references/phase-2-challenge.md. NOT a kill phase. Tight approach analysis (core loop, riskiest assumption, wedge, tech-stack fit, build estimate), then walk the student through scope decisions: app surface (Lean MVP / Standard / Full), connectivity (offline-only / offline-first / online), sharing (none / share-out / light social / full social), accounts, and a final core-value feature cut/add pass. For each scope-down decision, surface explicit Risk / Reward / Tradeoff bullets so the student is making the cut with eyes open. Each decision uses AskUserQuestion.
3. Phase 3 — Brief — read references/phase-3-brief-template.md. Emit plans/{slug}/idea-brief.md with refined-spec block + feature_seeds populated for kta-prd-pipeline and kta-design-spec consumption. Brief template has no monetization section — that's a separate pass.
4. Phase 4 — Handoff (MANDATORY) — after the brief file is written, use AskUserQuestion to offer the next step. Do NOT print raw text choices — use the tool. See "Handoff Prompt" section below.

Handoff Prompt (Phase 4)

Immediately after writing plans/{slug}/idea-brief.md, surface the absolute path of the brief in a one-line confirmation, then call AskUserQuestion:

header: "Next Step"
question: "Brief written to {abs path}. What next?"
multiSelect: false
options:
  - label: "Run PRD next", description: "Auto-run kta-prd-pipeline with this brief path"
  - label: "Exit to read", description: "Stop here so I can read the brief first"

Branching:

• If user picks "Run PRD next" → immediately invoke the kta-prd-pipeline skill via the Skill tool, passing the absolute brief path as input (e.g., args: "plans/{slug}/idea-brief.md"). Do NOT wait for further confirmation; the user already confirmed. The downstream skill takes over the conversation.
• If user picks "Exit to read" → end the skill cleanly. Print the brief path one more time so the user can open it. Do NOT proactively run any other skill.

This handoff is non-optional — every successful run of kta-idea-pipeline ends with this question. If the brief failed to write, skip the handoff and surface the error instead.

Inputs

• User intent (free text)
• Optional — niche, target platform, time/budget constraint

Outputs

• plans/{slug}/idea-brief.md — schema in references/phase-3-brief-template.md. Must include feature_seeds: (5–8 candidate features) for kta-prd-pipeline consumption.

Integration

• Next stage — kta-prd-pipeline (decomposes idea-brief into per-feature PRDs with screens nested)
• Out of scope here — monetization, pricing, billing, ads. Handle in a separate later pass once feature value is validated.
• Standalone — yes, output is human-readable

Constraints

• Phase 0 runs only when needed — don't force discovery when student already has an idea
• Phase 1 emits exactly 2 ideas — no more (token efficiency, easier comparison)
• Phase 1/2/3 NEVER ask about monetization, pricing, IAP, ads, subscription. If user volunteers monetization preferences, capture in a one-line note monetization_note: in the brief frontmatter and move on — no decision tree, no scope choice.
• Phase 2 is refine-not-kill — assume the idea ships, decide its shape via scope choices (offline/online, sharing, MVP vs full, accounts, core-value features). NO 1–5 scoring, NO kill verdicts.
• Phase 2 scope-down questions MUST surface Risk / Reward / Tradeoff bullets in the question description so the cut is reasoned, not reflexive.
• Skill emits markdown only, no code
• Default working directory — plans/{slug}/ (current plan's slug)
• WebFetch in Phase 0 is best-effort — has prompt-only fallback
• Phase 4 handoff is mandatory — every successful brief write ends with the AskUserQuestion offering "Run PRD next" (auto-chains to kta-prd-pipeline with brief path) or "Exit to read"