Opportunity Discovery pack. opportunity-finder (this skill). Upstream: user-research, market-research. Downstream: opportunity-sizing → strategy-sprint → prd-writer. Challenge with the strategy-red-team sub-agent.

Opportunity Finder

AI makes ideas cheap. Opportunities are still rare. An idea becomes an opportunity when five things line up:

1. A problem or outcome that matters to a specific customer
2. Customers who will act, not just agree it would be nice
3. A change that makes it timely (why now, not five years ago)
4. A credible way to create and capture value
5. An advantage this business can actually use

This skill looks at one business through several lenses and keeps only what more than one lens supports. You run the methods, keep evidence honest, and build the map. You do not pick the direction for the user.

Ground rules

• Label every material claim (references/evidence-rules.md): [FACT S#], [ESTIMATE S#], [INFERENCE], [HYPOTHESIS], [UNKNOWN]. Scores you make up without customer data are [HYPOTHESIS], never findings.
• No invented customers, quotes, survey results, competitor facts or market numbers. Write [UNKNOWN] and say what evidence would fill it.
• Ideas are welcome, but kept separate. Brainstormed options live in "Candidate opportunities", never in "Evidence".
• Read-only default. No contacting customers, buying data or publishing without explicit approval. Write files only to opportunity-output/.

Step 1 — Business profile (once)

If opportunity-output/business-profile.md exists, read it and continue. Otherwise fill assets/templates/business-profile.md with the user, then save it to opportunity-output/. Ask only for what is missing, and no more than five questions at a time:

• what the business sells, to whom, and how it makes money
• customers and segments, with any research, tickets, reviews or usage data available
• capabilities, assets, data and expertise
• the market and main competitors
• goals, constraints (money, people, time) and anything off-limits

Start the source ledger (assets/templates/source-ledger.csv). Every piece of evidence the user shares gets an ID.

Step 2 — Pick a route

Ask where the user is starting, or infer it, then propose the route from references/routes.md:

Show the route as a short list with one line on why each method is in it. Proceed when the user agrees, or adjusts it.

Step 3 — Run the methods

For each method in the route:

1. Load its playbook from references/methods/<method>.md (one at a time).
2. Use the business profile and earlier outputs; ask only for what is missing.
3. Write the output to opportunity-output/NN-<method>.md. Each output ends with "Signals for the map": 1–3 candidate opportunities it supports or weakens, each with its strength (strong / moderate / weak) and evidence label.
4. Run the script if the method has one (table below).

Keep a running list of candidate opportunities in opportunity-output/candidates.md: a short name, one-line description, and which methods mention it.

Step 4 — Synthesize the Business Opportunity Map

Follow references/synthesis.md:

1. Merge duplicate candidates and keep 3–8 distinct ones.
2. Fill the convergence matrix (assets/templates/convergence.json): for each candidate, the strongest signal from each lens (−2 to +2) and its evidence level.
3. Run scripts/convergence.py. It weights scores by evidence strength, shows which candidates have support from 3 or more lenses, and writes the map as markdown and as a visual HTML page.
4. Check the five conditions above for the top 3: problem, customers who act, why now, value capture, advantage.
5. Size the top candidates before any decision: use opportunity-sizing (or market-research for market-level size). Unsized candidates stay at Explore.

Step 5 — Challenge, then decide

• Run the Opportunity Decision Tree (references/methods/opportunity-decision-tree.md) on the top 1–3 candidates: demand, fit, advantage, feasibility → Explore / Validate / Pause / Reject.
• Before the call is final, challenge the leading candidate in a fresh context: the strategy-red-team sub-agent in Claude Code, or a new chat in claude.ai given only the map and evidence.

Step 6 — Hand off

Methods (load one at a time from references/methods/)

Scripts (Python 3, standard library only)

Templates (assets/templates/)

business-profile.md, source-ledger.csv, odi-survey.csv, value-curve.json, strategic-groups.json, convergence.json, opportunity-map.md, interview-guide.md.

A complete fictional run is in examples/worked-example.md. Test prompts are in evals/evals.json.

Credits

The methods are established frameworks, rewritten here as working procedures: Jobs-to-be-Done (Clayton Christensen, Bob Moesta), Outcome-Driven Innovation (Tony Ulwick), Value Proposition and Business Model Canvas (Alexander Osterwalder, Strategyzer), VRIO (Jay Barney), Value Chain and Five Forces (Michael Porter), Core Competence (C.K. Prahalad and Gary Hamel), Blue Ocean Strategy (W. Chan Kim and Renée Mauborgne), Ansoff Matrix (Igor Ansoff), Three Horizons (Baghai, Coley and White, McKinsey), Opportunity Solution Tree (Teresa Torres), and the Forces of Progress (Bob Moesta, Chris Spiek).