differentiated-opportunity-engine

name	differentiated-opportunity-engine
description	Identifies differentiated money-making opportunities for skilled solo operators, ranks them across cash-now, asset-later, and hybrid paths, and produces validation plans plus tool/build specs. Use when a user wants business ideas, monetization strategies, micro-SaaS wedges, service-to-software paths, or AI-leveraged opportunities matched to their skills, assets, access, credibility, constraints, and legal boundaries.
compatibility	Works best in environments with read/search tools and optional web access for market checks.
metadata	{"author":"opencode","version":"0.3"}

Differentiated Opportunity Engine

Overview

This skill finds money-making opportunities that fit the specific user instead of producing generic hustle lists. The core principle is to rank opportunities by founder edge first, then by speed to revenue, durability, and AI leverage, while being willing to say not yet when the user's constraints make the goal unrealistic.

Load references/scoring-rubric.md when scoring or comparing opportunities in detail. Load references/opportunity-patterns.md when the opportunity set needs widening. Load references/adversarial-checks.md when the user's preference, constraints, or legal situation make the obvious answer suspicious.

When to use

• When the user wants help finding business ideas, monetization paths, or side-income opportunities
• When the user wants AI-leveraged opportunities matched to their skills, assets, access, or network
• When the user wants to decide between services, micro-SaaS, automations, data products, or content-supported businesses
• When the user has multiple possible paths and needs a ranked recommendation with validation steps
• When the user wants a plan for how to exploit a niche advantage legally and practically

Do NOT use when:

• The user already chose an opportunity and only needs implementation details for a specific app, workflow, or campaign
• The request depends on illegal tactics, fraud, impersonation, IP theft, unlawful spam, or obvious platform abuse

Workflow

1. Build the differentiation profile

Extract the user's edge before brainstorming. Focus on skills, domain experience, assets, distribution, network access, credibility, capital/time constraints, and risk tolerance. Interests matter only as a tiebreaker unless the user makes them central.

If key data is missing, ask up to 5 short intake questions.

When the user's niche, buyer access, or distribution is weak or unclear, do not rank opportunities yet. First use 3-5 targeted questions to identify the narrowest credible wedge.

Only proceed with assumptions if the user cannot answer or explicitly wants a provisional answer. If you proceed without wedge clarity, mark the result as provisional and downgrade confidence.

Output a concise Differentiation Profile section with:

• Skills and domain experience
• Assets and access advantages
• Distribution and network leverage
• Constraints
• Credibility signals
• Unfair advantages

If distribution is weak, the intake must focus less on abstract interests and more on:

• who the user can reach now
• what repeated painful workflow they know best
• what prior work created trust or pull
• whether there is a clean-room-safe warm path
• whether a service-first wedge is acceptable

Recommendation gate for weak distribution:

If the user lacks a clear niche or owned distribution, do not jump directly from broad capability to a market recommendation. First identify a plausible initial wedge: a narrow buyer, one urgent workflow, one credible path to first conversations, and one legally usable trust asset.

If that cannot be identified after concise follow-up, the primary recommendation should usually be an edge-building or wedge-finding plan, not a fully specified business pick.

2. Run the feasibility and legality screen

Before ideating, test whether the user's requested outcome is credible under the current constraints.

Check for:

• Impossible bundles such as fast money + no selling + no content + no service + no audience + very little time
• Employer confidentiality, customer-data rights, IP restrictions, licensing, procurement conflicts, and regulated-workflow boundaries
• Whether the user's stated preference conflicts with the evidence

Important: if the user's only apparent edge comes from employer-confidential information, proprietary customer access, or regulated authority they do not legally control, treat that as no usable edge until proven otherwise.

Set a feasibility_verdict mentally as one of:

• go
• conditional
• no_go

If the right answer is conditional or no_go, say so plainly. Do not invent a glamorous path to satisfy the user's original preference.

3. Generate opportunity candidates

Create a broad but realistic set of candidates using references/opportunity-patterns.md. Always consider:

• Cash-now options
• Asset-later options
• Hybrid ladders such as service -> automation -> product

Also include at least one non-default candidate when it is relevant, such as:

• Audience or community-led products
• Data or benchmark subscriptions
• Templates or info products
• Partnerships, reseller, or licensing wedges
• No-go for now with an edge-building plan

When distribution is weak, always include at least one candidate explicitly optimized for that weakness, such as partner-channel, white-label, reseller, embedded specialist, or selling through someone who already owns buyer trust.

Prefer opportunities that can exploit the user's existing edge quickly. Do not default to pure content businesses. Treat content as a support layer unless the user has unusually strong distribution, audience trust, or direct demand capture.

Apply a platform-specificity penalty. Do not anchor on GCP, AWS, React, or any other ecosystem unless one of these is true:

• the user has independently owned buyer access in that ecosystem
• the platform creates a truly distinct pain
• early validation shows that the platform-specific version converts better than the platform-neutral version

Discard candidates that are generic, legally risky, highly platform-dependent, or impossible to validate with the user's current resources.

4. Score and rank opportunities

Score each candidate using the rubric in references/scoring-rubric.md. Every opportunity must answer:

Why does this user have an advantage over a random competent person with ChatGPT?

If that answer is weak, lower the rank even if the market is large.

If the answer depends mostly on generic skill or broad market familiarity, stop and ask follow-up instead of forcing a ranking.

If the user's preferred path scores poorly, include it anyway and explain why it loses.

For each serious candidate, include:

• Why this user can win
• Why this could fail
• Cash-now / asset-later / hybrid classification
• A short thesis
• A compact scorecard

Split the ranked list into:

• Cash Now
• Asset Later
• Hybrid Ladder

5. Recommend one path and pressure-test it

Pick the best near-term path, not the most glamorous one. Use this preference order only when the evidence is close:

1. Hybrid ladders
2. Strong-edge service or micro-SaaS plays
3. Content-supported businesses
4. Pure content/media plays only when the distribution edge is unusually strong

Pressure-test the top recommendation with:

• Demand clarity
• Distribution fit
• Execution difficulty
• Compliance and platform risk
• Whether a service-first wedge should come before software
• Why the obvious alternative loses
• What would change the recommendation

Also include one strongest alternative that tries to beat the winner. Do not use a token rejected path. State:

• why it might actually be better
• what assumption keeps the current winner ahead
• what evidence would flip the recommendation

If a service-first path improves odds of a later micro-SaaS, say that explicitly.

If you recommend service-first, specify all of these:

• A concrete first offer
• A narrow ICP
• One measurable pain or outcome metric
• A clear productization trigger

If you recommend content-led or audience-led monetization, justify it with the user's actual distribution strength and name monetization paths beyond ads.

If the correct answer is conditional or no_go, state the binding constraints, explain why the preferred path is not credible now, and give the shortest realistic path to improve the odds.

If the verdict is no_go, the main recommendation should usually be a constraint-change or edge-building plan, not a disguised product pick. You may mention the least-bad fallback path, but do not present it as realistic near- term income if the evidence does not support that.

6. Produce execution outputs

Always return both a human-readable report and a machine-readable JSON block. Use the template in assets/opportunity-analysis-template.json as the default shape.

Use these section headers by default unless a conditional or no_go answer needs a slight wording change:

• Differentiation Profile
• Top Opportunities
• Best Bet
• 7-Day Validation Plan
• Build Spec
• Kill Criteria

The JSON should capture the same result so later tools can consume it.

Requirements for substance, not just formatting:

• The best bet in prose and the best bet in JSON must match.
• The 7-day validation plan must include numeric targets where possible.
• Kill criteria must be falsifiable, preferably with counts, dates, or threshold conditions.
• Include at least one rejected path with a concrete reason.

Checklist

• The user's edge is explicit, not implied
• Feasibility, legality, and confidentiality have been screened
• Opportunities are split across cash-now, asset-later, and hybrid paths
• Every serious recommendation explains why this user can win
• The response explains why the best bet beats the obvious alternative
• If niche or distribution was weak, the skill asked wedge-finding questions before ranking opportunities
• The final recommendation names a narrow buyer, urgent workflow, and credible first channel or explicitly states that those are still unproven
• The best bet includes a 7-day validation plan, build spec, and kill criteria
• Validation targets and kill criteria are concrete, not generic
• The response includes both human-readable sections and a JSON block

Examples

Example:

Input: A staff data engineer with 8 years in logistics has relationships with 15 freight brokers, no audience, and 10 hours a week. They want a legal AI business and are open to services before software.

Output: A report that avoids generic AI content plays, ranks a broker-reporting automation service and a shipment-exception micro-SaaS highly, explains the user's logistics access advantage, recommends a hybrid ladder, says why a pure micro-SaaS-first path loses today, gives a 7-day validation plan with numeric targets based on broker interviews and a concierge pilot, includes falsifiable kill criteria, and ends with JSON containing the ranked opportunities and best bet.

{
  "user_profile": {
    "skills": ["data engineering", "workflow automation"],
    "network_access": ["15 freight brokers"],
    "constraints": ["10 hours/week"]
  },
  "feasibility_verdict": "go",
  "binding_constraints": [],
  "opportunities": [
    {
      "id": "opp_broker_reporting",
      "category": "hybrid",
      "title": "Broker reporting concierge to workflow SaaS",
      "why_user_can_win": "Existing freight relationships and domain fluency reduce discovery and sales friction.",
      "why_this_could_fail": "If brokers refuse to share workflow data or the problem is too bespoke to standardize.",
      "scores": {
        "edge_fit": 9,
        "speed_to_revenue": 8,
        "durability": 7,
        "distribution_fit": 8,
        "ai_leverage": 6,
        "defensibility": 7,
        "execution_difficulty": 5,
        "compliance_platform_risk": 2
      }
    }
  ],
  "rejected_paths": [
    {
      "title": "Generic freight AI dashboard",
      "reason": "Too broad and weakly differentiated for the user's current distribution."
    }
  ],
  "best_bet": "opp_broker_reporting"
}

Common mistakes

Mistake	Fix
Ranking generic AI wrappers highly	Penalize weak founder edge and weak distribution advantages
Defaulting to pure content/media businesses	Only keep content-first paths when the user's distribution edge is unusually strong
Treating interests as the main differentiator	Weight skills, assets, access, and credibility ahead of interests
Recommending software before validation	Prefer a service or concierge wedge when it improves learning speed
Giving ideas without kill criteria	State the signal that would cause the user to stop or pivot
Treating impossible constraints as compatible	Say the path is conditional or no-go and name what must change
Confusing domain familiarity with legal right to act	Screen for confidentiality, licensing, IP, and regulated-work boundaries

Quick reference

Operation	How
Profile the user	Extract skills, assets, access, credibility, constraints, then summarize them plainly
Run feasibility gate	Check constraint realism, legality, confidentiality, and whether the preference is credible
Score opportunities	Use references/scoring-rubric.md and rank by edge first, then revenue speed and durability
Expand candidates	Use references/opportunity-patterns.md to widen beyond the first obvious idea
Format output	Human-readable sections plus JSON using assets/opportunity-analysis-template.json

Key principles

1. Edge first — The skill exists to find opportunities the user can win, not attractive markets in the abstract.
2. Legal and practical — Reject illegal, deceptive, or obviously platform-abusive paths even if they look lucrative.
3. Validation before build — A fast paid pilot, concierge workflow, or manual test beats speculative product building.
4. Feasibility honesty — Say not yet when the user's constraint bundle or legal situation makes the desired path unrealistic.
5. Structured output — Always produce both decision-ready prose and machine-readable JSON so downstream tools can act on the result.

Optimization Notes

• Preserve the user's requested output shape exactly and do not substitute generic advice for concrete artifacts.
• Include exact commands, code structures, protocol fields, tags, parameters, file paths, or deliverable sections when the task asks for them.
• Make safety gates explicit before irreversible, destructive, externally visible, or compliance-sensitive actions.
• For multi-step work, present steps in execution order and include validation or rollback checks where relevant.
• Avoid overfitting to a single eval example: express lessons as reusable rules, not as task-specific answers.