discover

name	discover
description	Use after init-project to fill docs/product/start-project.md with AI research on competitors, monetization, valuation, and marketing.

Discover — Product Discovery

Overview

Fills sections 6, 8, 9, 10 (and 11 for iOS) of docs/product/start-project.md using AI-research and c-level-skills:* advisory skills. Runs sequentially so each section builds on the previous one — monetization references competitor pricing, valuation references unit economics, marketing references SOM constraints.

Orchestration: the section chain (Steps 4→5→6→7) is a genuine dependency pipeline — do not fan it out, the cross-references would break (_shared/references/orchestration-conventions.md → "NOT safe to parallelize"). The only safe parallelism here is the read-only MemPalace search batch in Step 2. Research stays on the c-level-skills (they pick their own model).

Type: Architect

When to use

• After /vladyslav:init-project — once sections 1-5 of docs/product/start-project.md are filled in manually
• Mid-project pivot — when the business model or target audience changes
• Before fundraising — for a consolidated discovery bundle

Prerequisites

• docs/product/start-project.md need not pre-exist — it is auto-created from the template if missing (see Step 1 "self-seed")
• Sections 1-5 (ідея, проблема, audience, MVP scope, non-goals) must be filled with real content — if any still contain <...> placeholders, that is a hard stop (Step 1)

Process

Step 1: Read context

Read:

• CLAUDE.md — project-level rules (derive project name from the # <name> title; fall back to the current directory basename)
• docs/product/prd.md (if exists)
• docs/product/start-project.md — full file, primary input for all sections

If docs/product/start-project.md is missing — self-seed it:

1. Resolve the plugin root: Glob ~/.claude/plugins/cache/vladyslav-marketplace/vladyslav/*/skills/discover/SKILL.md and take the directory two levels up. If Glob returns nothing, fall back to /Volumes/DevSSD/Development/vladyslav-skills.
2. Read <plugin-root>/skills/init-project/assets/StartProject.md.
   • If that template is also missing → stop with: "Template skills/init-project/assets/StartProject.md not found in the vladyslav-skills plugin. Please reinstall or run git pull."
3. Replace every occurrence of <PROJECT_NAME> in the template with the project name derived from CLAUDE.md (or directory basename).
4. Create docs/product/ if it does not exist, then write the substituted content to docs/product/start-project.md.
5. Inform the user: "Created docs/product/start-project.md from template. Sections 1–5 still contain placeholder text — please fill in your idea, problem statement, audience, MVP scope, and non-goals before the research sections are filled."
6. Continue with Step 2 below (do not stop — proceed into MemPalace search and section detection).

Hard stop (applies whether file existed or was just created):

• Sections 1-5 still contain <...> placeholders → stop, ask user to fill the foundation first

Step 2: MemPalace — batch search

Apply the verify-working-directory contract from <plugin>/skills/_shared/references/verify-pwd.md: confirms CLAUDE.md exists, derives the canonical MemPalace wing name, warns on stale-wing duplicates, and establishes the mandatory path-validation rule for the rest of this skill's MemPalace reads.

Run all searches upfront as parallel reads (independent, no shared state — dispatch in one batch):

mempalace_search wing=<wing> "competitors"
mempalace_search wing=<wing> "monetization pricing"
mempalace_search wing=<wing> "idea validation PMF"
mempalace_search wing=<wing> "marketing channels"

Cross-wing search by category keywords from section 1 (e.g. for a tax app → "tax accounting").

Surface hits — prior research for any dimension shouldn't be re-done.

Step 3: Detect scope

Use <plugin>/scripts/section-status.sh docs/product/start-project.md to get {filled: [...], pending: [...]} as JSON. The "skip done" mode auto-skips any section listed in filled.

Ask the user which sections to fill:

• All (default) — runs Steps 4 → 5 → 6 → 7 in order
• Custom — user picks specific steps
• Skip done — auto-detect which sections are already filled (non-empty, no <...> markers) and skip those

iOS auto-detection: run <plugin>/scripts/detect-stack.sh . and check .ios == true in the JSON output. If true → schedule Step 8 (Apple Check) after Step 7.

---

Per-section flow: Steps 4-7 below all follow the same shape — invoke c-level-skill(s), synthesize, overwrite the target section of start-project.md, write a MemPalace decision record. The shared flow plus per-step prompts, synthesis formats, and MemPalace templates live in <plugin>/skills/discover/references/discover-section.md. For each step below, read that reference's matching block and apply it.

Step 4: Competitive landscape → section 6

Apply the per-section flow from <plugin>/skills/discover/references/discover-section.md → "Step 4 — Competitive landscape (section 6)". Inputs: start-project.md sections 1, 2, 3, 7. Output: section 6 of start-project.md plus docs/product/competitors.md.

---

Step 5: Monetization → section 8

Apply the per-section flow from <plugin>/skills/discover/references/discover-section.md → "Step 5 — Monetization (section 8)". Prerequisite: section 6 should be filled. Inputs: start-project.md sections 1, 3, 6, 7. Output: section 8 plus optional docs/product/monetization.md.

---

Step 6: Idea validation → section 9

Apply the per-section flow from <plugin>/skills/discover/references/discover-section.md → "Step 6 — Idea validation (section 9)". Inputs: start-project.md sections 1, 2, 3, 4, 8. Output: section 9 plus docs/product/validation-plan.md (only if verdict is YELLOW or RED).

---

Step 7: Marketing hypothesis → section 10

Apply the per-section flow from <plugin>/skills/discover/references/discover-section.md → "Step 7 — Marketing hypothesis (section 10)". Inputs: start-project.md sections 1, 3, 8, 9. Output: section 10 plus optional docs/product/marketing-plan.md.

---

Step 8: Apple Check → section 11 (iOS only)

Skip if non-iOS project. If iOS detected (Step 3):

Invoke the vladyslav:discover-apple-check skill via the Skill tool. It handles section 11, docs/product/apple-review.md, and its own MemPalace record independently.

---

Step 9: Final synthesis

1. Re-run <plugin>/scripts/section-status.sh docs/product/start-project.md to confirm all targeted sections are now filled. Any path returning to pending is a regression to investigate.

   Re-read docs/product/start-project.md — confirm all targeted sections are filled

2. Identify contradictions (e.g. monetization assumes $20/mo but SOM from valuation can't support that price). Flag each contradiction to the user: reopen the relevant step or accept as "known tension"

3. Write docs/product/discovery-summary.md — 1-page executive view:

   • Idea + core thesis
   • Top 3 competitor threats + our edge
   • Pricing hypothesis + unit economics gist
   • Validation status (green/yellow/red) + top risk
   • Marketing channels + first-100 plan gist
   • Apple-check status (iOS only)

4. Write MemPalace milestone record: "Discovery pass complete for <project>, <date>. Key findings: <3 bullets>"

Step 10: Report

✓ Architect report — Discovery
- Steps run: <list>
- Steps skipped: <list + why>
- Contradictions found: <count>
- Validation verdict: <GREEN|YELLOW|RED>
- MemPalace records added: <count>
- Files updated:
  - docs/product/start-project.md (sections 6, 8, 9, 10[, 11])
  - docs/product/competitors.md
  - docs/product/discovery-summary.md
  - docs/product/monetization.md (if created)
  - docs/product/marketing-plan.md (if created)
  - docs/product/validation-plan.md (if yellow/red)
  - docs/product/apple-review.md (iOS only)

Next steps:
- /vladyslav:add-feature — start building once discovery verdict is GREEN
- /vladyslav:write-user-stories — document the validated feature set

Notes

• Never fabricate competitor data or market numbers. TBD and ROUGH are honest; invented figures are traps.
• A RED validation verdict is a gift — it saves months of building the wrong thing.
• Customer-dev questions must be action-intent ("When did you last try to solve X?"), not opinion ("Would you use an app?").
• Running discover twice refreshes research but won't duplicate MemPalace records (check_duplicate guards each write).