Product Thinking
Skill2 starsChanged 4 months agoMITView on GitHub
Reason from user problems and evidence of demand, not from features - for build/don't-build decisions, prioritization, MVP scoping, and interpreting feature requests and user feedback. Activate on "should we build", "users are asking for", "what should we prioritize", "scope the MVP", roadmap questions, and any feature request about to be taken literally. Trigger signal: a solution being evaluated before the problem it solves has been stated.
In this repository38 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- Adversarial VerifySkill
- Change ControlSkill
- Code ReconnaissanceSkill
- Correction PropagationSkill
- Debugging PlaybookSkill
- Deep DecompositionSkill
- Delegation DisciplineSkill
- Disclosure Is Not A FixSkill
- Discovery LoopSkill
- Divergent IdeationSkill
- Effort CalibrationSkill
- Empirical ValidationSkill
- Error RecoverySkill
- Extract ApproachSkill
- Failure Mode AwarenessSkill
- ForesightSkill
- Frontier Workflow ModeSkill
- Human HandoffSkill
- Intent ClaritySkill
- Leverage FirstSkill
- Live State TruthSkill
- Memory HygieneSkill
- Open MandateSkill
- Output StructuringSkill
- Plan GateSkill
- Proactive RigorSkill
- Prompt EngineeringSkill
- Publish HygieneSkill
- Research MethodologySkill
- Ruthless EditorSkill
- Scope FenceSkill
- Security Pattern ReviewSkill
- Self-Improvement LoopSkill
- Session OrientationSkill
- Structured ReasoningSkill
- Verification DisciplineSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
3 — Weigh the evidence of demand,: not the volume of the ask:
- Distinctive: line 62
Ask: "Enterprise customers are requesting SSO — should we build it?" Problem decode: IT admins, during procurement and offboarding, must …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Product Thinking
Purpose
Product decisions fail in a standard way: a feature gets built because it was requested, was buildable, and sounded good — and the underlying user problem was never stated, so nothing was actually solved and nobody can even tell. This skill enforces the decode: every feature is someone's proposed solution to an unstated problem, and the problem — plus real evidence anyone has it — is what decisions get made on.
When to use this skill
- Build/don't-build and prioritization decisions; roadmap tradeoffs.
- Interpreting feature requests and user feedback — especially specific, loud, or repeated asks.
- Scoping an MVP or first version; defining what "success" for a feature means.
- Evaluating product/business ideas (pairs with
structured-reasoningandfailure-mode-awareness).
When NOT to use this skill
- Executing an already-made product decision — building it well is
plan-gate+ siblings; re-litigating it isscope-fenceterritory. - Pure technical decisions with no user-behavior question in them.
- Contractual obligations ("client X paid for feature Y") — decode the problem anyway for design quality, but the build decision is made.
Operating procedure
1 — Restate the ask as a user problem. Template: [who], in [situation], is trying to [accomplish what], and is blocked/hurt by [what]. If you can't fill this from available evidence, that's the first finding — the request is a solution with no stated problem, and the next step is discovering whether one exists, not building.
2 — Decode the request, don't obey it. "Add an export-to-Excel button" might be: needs one number weekly (→ a scheduled email beats the button); needs data in another tool (→ integration/API); doesn't trust the dashboard (→ a trust problem no export fixes). The request constrains the solution space least; the problem constrains it usefully.
3 — Weigh the evidence of demand, not the volume of the ask:
- Requested by how many, and are they representative or just loud? n=3 vocal users is a signal to investigate, not a mandate to build.
- Distinguish "say they want" from "behavior shows they need" — workarounds in the wild (spreadsheets, scripts, support tickets) are stronger evidence than requests.
- Who would stop using / stop paying without it? That's the demand that matters, and it's usually a smaller set than the requesters.
4 — Cost it honestly. Effort includes: build + the maintenance annuity (every feature is maintained forever) + surface-area costs (docs, support, testing, UI complexity) + the roadmap item it displaces. "Small feature" almost always prices only the first term.
5 — Cut the smallest thing that tests the value hypothesis. An MVP is not a small version of the full plan — it's the cheapest artifact that answers whether the value is real. Ask: what could we remove and still learn that? Sometimes it's a concierge/manual version, a fake-door test, or fixing the top workaround — not software.
6 — Define success and kill criteria before building. A usage/outcome metric ("30% of weekly actives use it within a month") and a kill threshold ("under 5% → remove it"). Metrics defined after launch get defined as whatever happened.
7 — Keep "do nothing" and "non-product" on the table. Better docs, a manual process, a pricing change, or accepting the gap are legitimate outcomes of product reasoning — often the highest-ROI ones.
Quality bar
- The problem statement (step 1) exists and is evidenced — no build recommendation rests on an undecoded feature request.
- Demand evidence distinguishes loud from representative, and stated-want from demonstrated-need.
- The cost estimate includes maintenance and displacement, not just build time.
- Success and kill criteria are defined before any build recommendation.
Common failure modes
- Solution-in-search-of-problem: evaluating "should we build X" without ever stating what X solves and for whom. The most common and most expensive.
- Feature-request literalism: building exactly what the loudest customer described, solving their workflow's symptom and no one else's anything.
- The shrunken-clone MVP: a "minimal" version that's just the full vision with fewer features — big enough to cost real time, unfocused enough to test nothing.
- Vanity evidence: "everyone is asking for this" from three tickets; "users love it" from launch-week curiosity spikes.
- Free-feature pricing: ignoring the maintenance annuity and surface-area costs, so every marginal feature looks cheap and the product compounds into sludge.
- Post-hoc success: shipping first, then discovering the metric that makes it look good. Kill criteria defined in advance are the only honest kind.
Example
Ask: "Enterprise customers are requesting SSO — should we build it?" Problem decode: IT admins, during procurement and offboarding, must control access centrally and are blocked from purchasing at all without it (step 1 — fillable, strong). Evidence: 2 lost deals name it in writing (demonstrated need, not just stated want); current workaround is shared credentials — a security risk that pressures churn (behavioral evidence). Cost: build + ongoing per-IdP support burden + certification maintenance — not small, priced honestly. MVP cut: SAML with the one IdP both blocked deals use — tests willingness-to-close without building the whole matrix. Success: both stalled deals close within a quarter; kill/pause: if procurement still stalls on other requirements, SSO wasn't the real blocker — investigate before extending. Recommendation: build the narrow version, with the deals as the test.
Works with sibling skills
intent-clarityis the same decode instinct applied to this conversation's user; this skill applies it to the product's users — decode both, don't conflate them.structured-reasoning(tradeoffs, risk/reward) andfailure-mode-awareness(business catalog: demand assumed, unit economics) do the decision mechanics; this skill supplies the problem framing and demand evidence they run on.divergent-ideationwidens the solution space after the problem is decoded (step 2 hands it a real problem to ideate against).verification-disciplineguards the evidence claims ("2 lost deals" is checkable; "everyone wants it" is not).
Provenance and maintenance
Added 2026-07 in the expansion pass: product/build decisions recur across real work and were previously served by generic reasoning skills that lack the domain's core move — decoding requests into problems and grading demand evidence. Practices are standard product craft, not claims about any specific market. Re-verify by auditing recent build recommendations: any that can't produce their step-1 problem statement and pre-registered success metric were made without this skill firing.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

