agentleFS
toribryantoribryan/design-skills/skills/product-design-process/SKILL.md

product-design-process

Skill1 starsChanged 45 days agoMITView on GitHub

Runs a product design effort end to end - discover, define, design, validate, systematize, ship, and tell the story - by routing each phase to the right skill and reviewer and keeping a running project log. Use for "design this feature from scratch", "where do I start", "take this from idea to shipped", "plan the design process", or any request that spans more than one phase. Not for a single task that one skill already covers; load that skill directly.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read the repo's AGENTS.md, plans/ or docs/, and any Figma links before proposing work.

    line 33
  2. Distinctive:

    Place the work: Use the phase map below to decide where the work starts.

    line 40
  3. Very specific:

    Read-only subagents in agents/.

    line 80
  4. Distinctive:

    Every phase ends with a written decision and the evidence behind it.

    line 96
  5. Distinctive:

    If the project has a design system, new work uses it.

    line 101
  6. Distinctive:

    If the project has an AGENTS.md, its conventions win over any skill here.

    line 103
  7. Distinctive:

    Every skill in the collection.

    line 132

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Product design process

The router for the whole collection. It decides which phase the work is in, loads the skill for that phase, runs the reviewer that checks it, and records each decision so the next phase starts from what was learned instead of from scratch. It does not do the phase work itself.

When to use

  • A new feature, product, or redesign with no clear starting point
  • Work that crosses phases ("research this, then design it, then build it")
  • Picking the work back up after a break and needing to know what comes next
  • Planning a design effort's steps and timeline

Not for: one bounded task ("critique this screen", "make motion tokens"). Load that skill directly.

Inputs

  • The ask, in the requester's words
  • What already exists: research, briefs, designs, a design system, code. Read the repo's AGENTS.md, plans/ or docs/, and any Figma links before proposing work. [Assume nothing exists.]
  • Appetite: how much time this deserves (a day, a week, six weeks). [One week.]
  • Fidelity needed at the end: a decision, a prototype, shipped code. [Shipped code.]

Process

  1. Place the work. Use the phase map below to decide where the work starts. Most requests arrive in the middle: "design a settings page" is phase 3 with phase 1 and 2 questions still open. Name the skipped phases and the risk of skipping each in one line.
  2. Open a project log. Model: templates/project-log.md. Copy that shape. Keep it in the repo (plans/NNN-<name>.md if the repo has a plans/ folder, otherwise docs/design/<name>.md).
  3. Run each phase. Load the phase's skill, produce its output, then run its exit check (below). Record the decision and the evidence in the log before moving on.
  4. Size to the appetite. With a day, run phase 1 as a 30-minute framing pass and phase 4 as a design-critic pass. With six weeks, run every phase fully. Never drop Verify.
  5. Loop back when evidence says so. A usability finding with severity 3 or 4 sends the work back to phase 3. A finding that the problem was wrong sends it back to phase 1. Record the loop and why.
  6. Close. Ship, write the changelog line, and, if the work is portfolio-worthy, hand the log to design-case-study while it is fresh.

Phase map

PhaseQuestion it answersSkillsExit check
1. DiscoverWhat problem, for whom, and what do we not know?design-discovery, user-research, reference-researchProblem statement agreed; riskiest assumption named and tested or accepted
2. DefineWhat will we build, how will we know it worked?design-brief, information-architectureBrief with success metrics and non-goals; flows and state inventory cover the edge cases
3. DesignWhat does it look like and how does it behave?wireframing, layout-and-hierarchy, ux-writing, interface-polishThree directions explored before one was chosen; all states drawn; design-critic verdict ready
4. ValidateDoes it work for real people?usability-testing, design-critique, accessibility-reviewNo open severity 3–4 findings; accessibility-auditor verdict ready
5. SystematizeWhat should become reusable?design-tokens, color-system, typography-system, iconography, component-api-design, design-system-docs, design-system-auditNew parts documented; token-auditor and component-reviewer verdicts ready
6. ShipIs what shipped what was designed?design-handoff, design-to-code, visual-qavisual-qa-reviewer verdict ready; project checks pass
7. TellWhat did we learn, and how do we show it?design-case-studyOutcome measured against the brief's metrics

Brand, motion, and marketing run alongside the phases, not in sequence:

TrackSkillsWhen it enters
Brandbrand-strategy → brand-voice → brand-identity → art-directionBefore phase 3 for a new product or rebrand; otherwise read the existing guidelines at phase 3
Motionmotion-language → motion-implementation → micro-interactionsLanguage at phase 5, implementation at phase 6, micro-interactions during phase 3 polish
Marketinglanding-page-designAfter brand and art direction exist

Reviewers

Read-only subagents in agents/. Each reports failures only and ends in a verdict. Run them at the exit of the phase they guard, and run them again after fixes.

ReviewerGuards
design-criticPhase 3 and 4
accessibility-auditorPhase 4 and 6
copy-reviewerPhase 3 and 6
component-reviewerPhase 5
token-auditorPhase 5 and 6
visual-qa-reviewerPhase 6
brand-reviewerBrand track and marketing

Standards

  • Every phase ends with a written decision and the evidence behind it. A phase with no record did not happen.
  • Explore at least three genuinely different directions before converging, at whatever fidelity the appetite allows.
  • Carry the brief's success metrics through to the end; phase 7 measures them.
  • If the project has a design system, new work uses it. New parts go through phase 5 before phase 6, not after.
  • If the project has an AGENTS.md, its conventions win over any skill here.

Output

The project log, filled in as the work moves, plus each phase's deliverable. At any point, the log answers: where are we, what was decided, why, and what is next.

Verify

The effort is done when:

  • Every phase that ran has a decision and evidence in the log
  • Every skipped phase is named with its accepted risk
  • All reviewer verdicts for the phases that ran read ready
  • The shipped result was checked in light and dark, by keyboard, with reduced motion on, and at 320px and 1440px
  • The brief's success metrics have a date when they will be measured

Anti-patterns

  • Jumping to high fidelity on day one because the tool makes it easy
  • Research that produces a deck and no decision
  • One direction presented as if it were the only option
  • A design system built before two real screens needed the same part
  • Treating accessibility and QA as phase 7 cleanup
  • A case study written from memory months later

Related skills

Every skill in the collection. This one only routes.

References

  • REFERENCE-BANK.md: Nielsen Norman Group for method choice at each phase; Laws of UX for the principles phase 3 leans on.
  • FIBO.md: Fibo's plans/ folder is the model for the project log (numbered, four questions, kept after shipping), and its definition of done is the model for phase 6's exit check.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.