product-design-router
Skill4 starsChanged 20 days agoMITView on GitHub
Diagnose the highest-risk unknown in product design or independent product-building work, establish decision ownership, and compose the smallest useful skill loop. Use when the user is unsure what work should happen next, has mixed artifacts from different stages or surfaces, or wants to avoid mechanically following a lifecycle. Do not use to replace a clearly requested specialist skill.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- Gaurav890/everything-agentic-engineeringtheir own version, first seen Sep 13, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- Gaurav890/everything-agentic-engineering★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository9 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 10 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈10
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 16
Before choosing a skill, make the smallest shared working state explicit:
- Distinctive: line 24
Do not invent an organization chart.
- Distinctive: line 37
Prefer the unknown whose answer would most change the product, not the one with the easiest artifact.
- Distinctive: line 43
Use references/CAPABILITY-NETWORK.md only when choosing among multiple nodes or composing a cross-node loop.
- Distinctive: line 79
Do not present a complete skill catalog when one next action is enough.
- Distinctive: line 80
Do not treat a polished PRD, mock, prototype, or codebase as proof that upstream decisions are settled.
- Distinctive: line 81
Do not invent research, analytics, constraints, or user consent.
- Distinctive: line 82
Do not optimize for artifact completion.
- Distinctive: line 83
Do not treat a shared canvas or real-time generation as shared understanding.
- Distinctive: line 84
Do not assign AI final authority over ambiguous, high-impact, or irreversible product decisions.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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 Design Router
Route by uncertainty, not by lifecycle stage.
Start from what exists
Inspect the supplied brief, research, prototype, code, analytics, feedback, constraints, and decision records. Do not require discovery artifacts merely because they traditionally come first. If the user has already named a concrete specialist task, preserve that intent and invoke that skill directly unless a missing decision would make the work misleading or unsafe.
Establish the collaboration and decision spine
Before choosing a skill, make the smallest shared working state explicit:
- the current working artifact or source of truth, including its version when relevant;
- the product decision and the person accountable for making or approving it;
- the people and agents contributing evidence, generation, critique, or implementation;
- the working surface where the uncertainty can best become observable: source evidence, canvas, prototype, code, or a live product signal;
- what AI may propose or execute and what requires human review or approval.
Do not invent an organization chart. One person may own several responsibilities in an independent-builder context. Record only the ownership and authority boundaries needed for the current decision.
Diagnose the decision
- Name the product decision the user is trying to make.
- Separate known evidence, inference, assumption, preference, and confirmed decision.
- Identify the unknown most likely to cause failure or expensive rework:
Value: wrong user, buyer, problem, outcome, or alternative;Meaning & Structure: incoherent objects, relationships, states, rules, IA, or permissions;Experience & Judgment: unclear interaction, content, usability, accessibility, or quality bar;Feasibility & Reliability: implementation, performance, safety, recovery, or operational risk;Adoption & Market: weak positioning, channel, activation, pricing, retention, or distribution;Learning & Governance: missing instrumentation, interpretation, provenance, approval, or decision memory.
- Prefer the unknown whose answer would most change the product, not the one with the easiest artifact.
- Choose the working surface that exposes the unknown most directly. Do not default to a canvas, document, prototype, or code merely because it is convenient or fast.
- Propose the smallest loop that can produce credible evidence. Usually select one primary skill and at most two supporting skills.
- Assign only the collaboration needed for that loop: who supplies context, who creates or changes the artifact, who evaluates it, and who owns the decision.
- Define an exit condition: what evidence would support, reject, or reframe the current decision.
Use references/CAPABILITY-NETWORK.md only when choosing among multiple nodes or composing a cross-node loop.
Output
## Entry state
- Working artifact / source of truth:
- Current working surface:
## Decision now
## Highest-risk unknown
- Node:
- Why now:
- Evidence already available:
- Missing evidence:
## Collaboration & decision spine
- Decision owner:
- Contributors and agents:
- AI may:
- Human review / approval:
## Smallest useful loop
1. Primary skill — question it should answer
2. Optional supporting skill — only if required
## Exit condition
## After the loop
- Update:
- Next unknown, if any:Boundaries
- Do not present a complete skill catalog when one next action is enough.
- Do not treat a polished PRD, mock, prototype, or codebase as proof that upstream decisions are settled.
- Do not invent research, analytics, constraints, or user consent.
- Do not optimize for artifact completion; optimize for evidence that can change a decision.
- Do not treat a shared canvas or real-time generation as shared understanding; make the decision, evidence, and owner explicit.
- Do not assign AI final authority over ambiguous, high-impact, or irreversible product decisions.
- High-risk legal, medical, financial, security, privacy, or irreversible decisions require the relevant specialist review and human approval.
- Designer and Builder profiles adjust default emphasis only; they never force a fixed sequence.
The route is complete when the user can see why this unknown matters now, where the work should happen, which minimal capability to invoke, who owns the decision, and what new evidence will end the loop.
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.

