{"id":53813,"plugin_id":"plugins_6ac5e934d1f88191b73fe472c16744ae","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-10-24T13:55:10.095Z","digest":"2057e2d251e928aeef8b0cc707048250f0e3729a9046365627e52943354123d6","against":null,"payload":{"description":"Plan qualitative customer discovery and synthesize real interview, observation, and first-party customer evidence without inventing or overstating findings.","included_files":[],"name":"customer-discovery","skill_md_contents":"---\nname: customer-discovery\ndescription: Plan qualitative customer discovery and synthesize real interview, observation, and first-party customer evidence without inventing or overstating findings.\n---\n# Customer Discovery\n\nUse this workflow to learn how a defined group experiences a problem, what they do now, how they decide, and what evidence would change a business decision. It supports both planning research and analyzing real material the user supplies.\n\n## Boundary with market research\n- Use `market-research` for secondary research on markets, categories, competitors, and current external facts.\n- Use `customer-discovery` for primary research planning and synthesis of real interviews, observations, customer conversations, and first-party behavior.\n- Use both when the decision needs both market context and direct customer evidence. Neither alone guarantees demand or business viability.\n- The AIOS can prepare a study and synthesize supplied evidence. It must not imply that it recruited, interviewed, observed, surveyed, or contacted people unless that actually happened through an available, authorized tool.\n\n## Mode A — Plan discovery\n1. **Define the decision.** State the business decision the research should inform, the current assumption, what is already known, and the most important unknown. Keep one study focused on a small number of related questions.\n2. **Define who can inform it.** Specify relevant experience or behavior for participants. Avoid treating broad demographics as proof of fit. Identify whose experience might contradict the leading assumption, including people who chose a substitute, stopped using a solution, or decided not to act.\n3. **Choose a suitable method.** Recommend interviews, contextual observation, workflow review, or a small exploratory survey only when appropriate. Explain what the method can and cannot establish. Qualitative discovery can reveal patterns and explanations; a small convenience sample cannot estimate population prevalence.\n4. **Prepare ethical recruitment and consent.** Provide a transparent invitation and screening criteria. Make participation voluntary, explain the purpose and intended use of notes/recordings, request consent before recording or quoting, avoid unnecessary sensitive personal information, and advise secure handling and de-identification.\n5. **Write a neutral discussion guide.** Ask about recent, concrete past situations before hypothetical solutions: what happened, what they did, what alternatives they considered, who was involved, what made the situation harder or easier, what it cost in time/money/effort when they know, and what happened next. Ask open questions, avoid leading wording and praise, and do not treat “Would you buy this?” or a compliment as purchase evidence.\n6. **Provide a note template.** Include a participant code, date, recruitment context, relevant screening detail, consent/recording status, question/topic, accurate verbatim quote versus paraphrase, observed behavior versus interpretation, and follow-up question. Do not add plausible-sounding quotes.\n\n## Mode B — Synthesize supplied evidence\n1. Confirm what source material exists and its provenance: real interview notes, verbatim transcript, observation, survey response, customer record, user recollection, or synthetic example. If source and scope are missing, ask or label the limitation.\n2. Preserve participant identifiers as anonymous codes. Separate verbatim quotes from paraphrases, observations from interpretations, and researcher notes from what participants actually said or did.\n3. Code recurring topics without forcing every case into the same pattern. Track supporting cases, counterexamples, disagreement, differences by context, current workarounds, decision factors, consequences, and unresolved explanations.\n4. Distinguish a repeated theme in this sample from prevalence in a wider market. Never invent counts, percentages, saturation, participant agreement, or causal claims. When counts are calculated from supplied data, state the denominator and sample scope.\n5. Compare findings with the starting hypothesis and seek disconfirming evidence. Report what is supported, weakened, contradicted, or still unknown. Treat stated interest separately from observed action, commitment, and willingness to pay.\n6. Recommend the next smallest useful research step or business experiment. Do not turn exploratory findings directly into a validated ICP, offer, price, or launch decision.\n\n## Evidence labels\n- **USER-SUPPLIED PRIMARY MATERIAL:** raw notes, transcripts, observations, or first-party records supplied by the user and represented as real. Analyze them as primary material within the stated scope, but attribute them as user-supplied; the AIOS has not independently authenticated their origin, completeness, participant count, consent, or collection method unless that was separately verified.\n- **USER-REPORTED SUMMARY:** the user says research occurred or summarizes findings, but the underlying material was not directly inspected. Keep it attributed and do not upgrade it to inspected primary material or independently verified evidence.\n- **CORROBORATED PRIMARY EVIDENCE:** use only when the source and relevant collection facts have actually been independently corroborated; state what was checked and what was not.\n- **EXAMPLE / SYNTHETIC:** invented or simulated material used to test a question, format, or reasoning process. It is never customer evidence and cannot support market or buying claims.\n- **HYPOTHESIS:** an interpretation or proposition that requires further validation.\n- **UNKNOWN:** a material question the available evidence does not answer.\n\nIn every synthesis, report source provenance and verification limits. Do not call user-supplied material independently verified merely because it is formatted as interview notes or a transcript. If the user asks for simulated interviews, fictional responses, or a composite persona, label every such item **EXAMPLE / SYNTHETIC — NOT CUSTOMER EVIDENCE** in the output. Never write fabricated quotations or stories as if they came from research.\n\n## Standard output\nFor a study plan: decision, research question, hypothesis, participant profile, recruitment/screening, method, consent/privacy notes, neutral guide, note template, limitations, and next step.\n\nFor synthesis: evidence inventory and provenance, sample/context, themes with supporting and conflicting cases, accurate quotes only when supplied, alternative explanations, what the evidence does/does not establish, revised hypotheses, unknowns, and next research step.\n\n## Candid partnership and research standard\nBe a candid strategic partner, not an echo chamber or cheerleader. Do not agree or praise merely to please. Challenge material assumptions, weak reasoning, gaps, contradictions, and meaningful risks; explain why they matter and offer a better-supported alternative when possible. Agree when the reasoning and evidence support it; do not invent objections or debate subjective preferences as facts.\n\nFor material business, customer, market, competitor, demand, pricing, performance, platform, regulatory, or feasibility claims, make relevant research and evidence the primary basis. Prefer direct/first-party evidence and current authoritative sources appropriate to the question; cite and scope sources, look for counterevidence, and state limitations. User or project preferences guide choices but do not prove factual claims. If evidence is missing or mixed, say so and keep the conclusion HYPOTHESIS or UNKNOWN. Purely creative work needs no unrelated research; factual claims in creative assets still need support.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}