ln-201-opportunity-discoverer（ln 201 机会发现者）

Paths: File paths (shared/, references/, ../ln-*) are relative to skills repo root. If not found at CWD, locate this SKILL.md directory and go up one level for repo root. If shared/ is missing, fetch files via WebFetch from https://raw.githubusercontent.com/levnikolaevich/claude-code-skills/master/skills/{path}.

Opportunity Discoverer

Type: L3 Worker Category: 2XX Planning

Traffic-First approach to finding next growth direction for existing product.

Core Philosophy

Anti-pattern: Idea → Surveys → Product → "where's traffic?" Correct: Traffic → Niche → MVP → Launch under existing demand

The 90% Developer Bug

Most fail because they:

1. Invent idea with no analogs
2. Ask 5 people "would you pay?" (they say yes for a hot dog)
3. Build product with round sum
4. Launch with "now let's set up traffic"
5. Discover: no traffic exists, never did

No marketer will build funnel for what cold traffic doesn't buy.

Traffic-First Principles

Supporting Methodology

Marc Andreessen (pmarca):

"Validate market at practical level — go get paying customers to demonstrate market exists."

Sam Altman (YC):

"Who desperately needs the product? Best answer is going after large part of small market." "Test idea by launching or trying to sell — get letter of intent before code."

Purpose & Scope

• Discover growth direction BEFORE Epic creation
• Filter ideas through evidence-first KILL funnel
• Output: one recommended idea + one traffic channel
• Position: before ln-210 (Epic Coordinator)

Runtime Contract

MANDATORY READ: Load shared/references/planning_worker_runtime_contract.md, shared/references/coordinator_summary_contract.md

Runtime family: planning-worker-runtime

Identifier:

• discovery work item identifier

Phases:

1. PHASE_0_CONFIG
2. PHASE_1_INPUT_PROCESSING
3. PHASE_2_KILL_FUNNEL
4. PHASE_3_RANK_SURVIVORS
5. PHASE_4_WRITE_DISCOVERY_REPORT
6. PHASE_5_WRITE_SUMMARY
7. PHASE_6_SELF_CHECK

Summary contract:

• summary_kind=opportunity-discovery-worker
• standalone mode may return the summary without artifact persistence
• managed mode writes the same JSON to summaryArtifactPath
• default managed artifact path pattern: .hex-skills/runtime-artifacts/runs/{parent_run_id}/opportunity-discovery-worker/ln-201--{identifier}.json

When to Use

Use this skill when:

• Product exists, seeking next growth direction
• Have 3-10 potential ideas/niches
• Want to validate opportunity before committing
• Need to choose ONE channel to focus on

Do NOT use when:

• No product context (greenfield startup)
• Already have validated direction (skip to ln-210)
• Prioritizing existing Stories (use ln-230)

Input Parameters

Input modes:

• ideas="idea1, idea2, idea3" — evaluate list
• context="SaaS for X" — generate ideas from product
• Both — generate + add user ideas

KILL Funnel Pipeline

Ideas do not go through 4 separate research-heavy passes anymore. Each idea gets one bundled evidence pass first.

Idea → [Evidence bundle: traffic + demand + competition + revenue]
          ↓
      [Hard kill matrix]
          ↓
     [Interest gate]
          ↓
      [MVP gate]
          ↓
       SURVIVOR

Evidence Bundle (single research pass)

Question: Is there enough external evidence to justify deeper evaluation?

Research bundle:

WebSearch: "[idea] how people find solutions"
WebSearch: "[idea] search volume {current_year}"
WebSearch: "[idea] competitors {current_year}"
WebSearch: "[idea] pricing SaaS"

Extract four signals in one pass:

• Traffic channel: Where do people actively look for this solution?
• Demand: Search volume, trend direction, or strong community pain signal
• Competition: Competitor count and ocean type
• Revenue: Plausible price band and willingness to pay pattern

Traffic channel examples:

Demand thresholds:

Competition thresholds:

Revenue thresholds:

Hard Kill Matrix

Kill immediately when any hard-stop condition is true:

• no identifiable traffic channel
• demand clearly below viable threshold with no compensating niche signal
• competition index = 5 and no clear wedge
• expected revenue below $20/user for a small-team business

Record the kill reason and stop analysis for that idea.

Personal Interest

Question: Will you enjoy building this?

Method: AskUserQuestion — rate 1-5

Rate your interest in building [idea]:
1 = Meh, would do for money only
2 = Low interest
3 = Neutral
4 = Interested
5 = Excited, would build for free

Why this matters:

• Low interest = burnout in 3 months
• High interest = sustained motivation through hard times
• You'll spend 2+ years on this

When to ask: Only for ideas that survive the external evidence bundle.

KILL if: Score 1-2 — you'll quit before PMF.

Output: Score 1-5

MVP-ability

Question: Can you launch in 4 weeks?

Assessment:

Time estimates:

When to assess: Only for ideas that survive external evidence + interest gate.

KILL if: >8 weeks to MVP — too slow to validate.

Output: Weeks estimate + blockers

Workflow

Phase 1: Input Processing (2 min)

1. Parse input:

- If ideas: split comma-separated list - If context: generate 5-7 ideas via WebSearch - If both: combine

1. Validate count:

- Minimum: 3 ideas - Maximum: 10 ideas

1. Create output directory: mkdir -p docs/reference/research/

Output: Idea queue (3-10 items) and checkpoint for PHASE_1_INPUT_PROCESSING

Phase 2: KILL Funnel (per idea)

Process each idea through one bundled evidence pass, then the personal filters only for survivors:

FOR each idea:
    Build evidence bundle:
        traffic + demand + competition + revenue

    Apply hard kill matrix
        IF failed → KILL, log reason, NEXT idea

    Ask Interest
        IF score 1-2 → KILL, log reason, NEXT idea

    Assess MVP-ability
        IF >8 weeks → KILL, log reason, NEXT idea

    → SURVIVOR: add to survivors list

Token efficiency:

• Process ONE idea at a time
• One research bundle per idea instead of four separate research phases
• KILL early = no interest prompt, no MVP assessment
• Clear context after each idea

Phase 3: Rank Survivors (2 min)

If survivors exist:

1. Calculate composite score: Score = Demand_score + (6 - Competition_index) + Revenue_score + Interest + MVP_score
2. Sort by score descending
3. Select TOP recommendation

If no survivors:

• Report: "All ideas killed. Rethink direction."
• Show KILL log for learning

Phase 4: Output (2 min)

Generate: docs/reference/research/[YYYY-MM-DD]-discovery.md

Also emit structured runtime summary:

• schema_version
• summary_kind=opportunity-discovery-worker
• run_id
• identifier
• producer_skill=ln-201
• produced_at
• payload with input_mode, ideas_analyzed, generated_ideas, survivors_count, killed_count, top_recommendation, report_path, warnings

Structure:

# Opportunity Discovery: [Date]

## Summary
- Ideas analyzed: X
- Survivors: Y
- Killed: Z

## TOP RECOMMENDATION

**Idea:** [Name]
**Channel:** [Primary channel]
**Why:** [2-3 sentence rationale]

### Key metrics:
- Demand: [volume]/month
- Competition: [Index] [Ocean type]
- Revenue: $[X]/user
- MVP: [X] weeks

## Survivors Table

| Idea | Channel | Demand | Competition | Revenue | Interest | MVP | Score |
|------|---------|--------|-------------|---------|----------|-----|-------|
| ... | ... | ... | ... | ... | ... | ... | ... |

## KILL Log

| Idea | Killed at | Reason |
|------|-----------|--------|
| ... | ... | ... |

## Next Steps
1. Create Epic with ln-210 for top recommendation
2. Focus on [channel] as primary acquisition
3. Target MVP in [X] weeks

Time-Box

Note: KILL funnel is faster than full scoring — bad ideas die early.

Integration

Position in workflow:

Product exists
     ↓
ln-201 (Opportunity Discovery) ← THIS SKILL
     ↓
ln-210 (Epic Coordinator)
     ↓
ln-220 (Story Coordinator)

Dependencies:

• WebSearch (all filters except Interest)
• AskUserQuestion (Interest filter)
• Write, Bash (output)

Critical Rules

1. Traffic first — no traffic channel = no analysis
2. Bundle evidence once — do not run separate research-heavy phases if one pass can answer traffic, demand, competition, and revenue
3. KILL immediately — don't score dead ideas
4. One recommendation — avoid paralysis
5. No surveys — real search data only
6. Interest matters — ask only for externally viable ideas
7. MVP speed — slow launch = slow learning

Example Usage

With ideas:

ln-201-opportunity-discoverer ideas="AI writing tool, code review bot, translation API"

With context:

ln-201-opportunity-discoverer context="B2B developer tools SaaS"

Example output:

# Opportunity Discovery: 2026-01-29

## TOP RECOMMENDATION

**Idea:** Code review bot
**Channel:** SEO (developers search "code review tool")
**Why:** Growing demand (15K/mo), emerging market (3 competitors),
$50/user pricing proven, can MVP in 3 weeks with existing skills.

## KILL Log

| Idea | Killed at | Reason |
|------|-----------|--------|
| AI writing | Competition | Red Ocean (25+ competitors) |
| Translation API | Revenue | Commoditized, <$10/user |

Definition of Done

• Ideas brainstormed from product context and market signals
• Evidence bundle collected for each idea before kill decisions
• Hard kill matrix applied before interest and MVP checks
• Survivors scored and ranked
• Discovery document generated at docs/reference/research/[YYYY-MM-DD]-discovery.md
• TOP RECOMMENDATION identified with channel + rationale
• KILL Log documents all eliminated ideas with reasons
• Structured opportunity-discovery-worker summary returned
• Summary artifact written when summaryArtifactPath is provided

Reference Files

• MANDATORY READ: Load shared/references/research_tool_fallback.md

Version: 2.0.0 Last Updated: 2026-01-29