design
Skill11 starsChanged 11 days agoMITView on GitHub
Defines product requirements from validated ideas, then produces detailed developer-ready specs. Use when the user says "design this", "write the PRD", "turn my prototype into requirements", "write user stories", "review the design", "design a prototype", or invokes /design. Covers prototype-first entry (static HTML/Tailwind CDN or Next.js/React/shadcn component battery), plus create, edit, and validate modes; routes the prototype track and requires designer–tech collaboration before handoff.
198 other projects use a version of this.
Adopted by
Most recently taken up by
- irank786/Claude-skillstheir own version, first seen Nov 16, 2026
- GoBeromsu/craft-skillstheir own version, first seen Nov 16, 2026
- david33137/claude-skillstheir own version, first seen Nov 16, 2026
- StitchSlop/claude-plugintheir own version, first seen Nov 16, 2026
- FAHTHIZ/oasistheir own version, first seen Nov 16, 2026
- fatihkan/baditheir own version, first seen Nov 16, 2026
Used with
Projects with it also use banner-design, ui-styling and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
lithStudySkillbanner-design
Design banners for social media, ads, website heroes, creative assets, and print. Multiple art direction options with AI-generated visuals. Actions: design, create, generate banner. Platforms: Facebook, Twitter/X, LinkedIn, YouTube, Instagram, Google Display, website hero, print. Styles: minimalist, gradient, bold typography, photo-based, illustrated, geometric, retro, glassmorphism, 3D, neon, duotone, editorial, collage. Uses ui-ux-pro-max, frontend-design, ai-artist, ai-multimodal skills.
Also in 39 projects that use this
MicrosoftSkillui-styling
Guide for the visual style, structure, and shared building blocks used by Semantic Link Labs interactive UI tools (HTML widgets and anywidget-based widgets). Use this when adding a new interactive UI, modifying an existing one, or adding shared visual components.
Also in 41 projects that use this
lithStudySkillbrand
Brand voice, visual identity, messaging frameworks, asset management, brand consistency. Activate for branded content, tone of voice, marketing assets, brand compliance, style guides.
Also in 44 projects that use this
OpenAnalystIncSkillads-management
Activate for paid advertising campaigns on Google Ads, Meta Ads, LinkedIn Ads, TikTok Ads. Includes ad copywriting, audience targeting, budget optimization, A/B testing, and ROAS tracking. Used by ads-specialist and campaign-manager agents.
Also in 3 projects that use this
OpenAnalystIncSkillai-artist
Write and optimize prompts for AI-generated outcomes across text and image models. Use when crafting prompts for LLMs (Claude, GPT, Gemini), image generators (Midjourney, DALL-E, Stable Diffusion, Imagen, Flux), or video generators (Veo, Runway). Covers prompt structure, style keywords, negative prompts, chain-of-thought, few-shot examples, iterative refinement, and domain-specific patterns for marketing, code, and creative writing.
Also in 3 projects that use this
mayankjain1815-commitsSkillinfra-audit
Reads docker-compose, env files, ORM configs, and connection strings to map current infrastructure. Flags missing layers (cache, queue, analytics) based on observed access patterns. Outputs a structured infrastructure manifest.
Also in 3 projects that use this
- +5 more from OpenAnalystInc
Other versions198 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 198
- nextlevelbuilder/ui-ux-pro-max-skill★ 150ksame name
- Yeachan-Heo/oh-my-codex★ 34ksame name
- tokio-rs/toasty★ 3.3ksame name
- PaulWoitaschek/Voice★ 3.3ksame name
- joye61/pic-smaller★ 2.9ksame name
- oxsecurity/megalinter★ 2.7ksame name
- Infisical/agent-vault★ 2.5ksame name
- static-web-server/static-web-server★ 2.4ksame name
- SethGammon/Citadel★ 927same name
- Ohh-889/skyroc-admin★ 795same name
188 more not listed
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 repository44 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- analyze-dataSkill
- brainstormingSkill
- craft-lessonSkill
- create-agentSkill
- create-skillSkill
- customize-agentSkill
- customize-skillSkill
- deslopSkill
- developSkill
- discovery-reportSkill
- elicitationSkill
- empathy-mapSkill
- explore-ideaSkill
- grill-meSkill
- help-meSkill
- humanizeSkill
- incidentSkill
- iterateSkill
- journey-mapSkill
- jtbdSkill
- kanbanSkill
- launchSkill
- legal-safety-auditSkill
- maintenance-executeSkill
- maintenance-planSkill
- memorySkill
- motionSkill
- phaseSkill
- planSkill
- research-planSkill
- retroSkill
- reviewSkill
- root-causeSkill
- round-tableSkill
- shape-upSkill
- shut-upSkill
- statusSkill
- teach-meSkill
- testSkill
- unpack-problemSkill
- validate-ideaSkill
- validation-patternsSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. 0 are common, 5 are distinctive, and 8 are very specific to this project.
0
Common
Most of the 175 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 21
No brief and no prototype exist yet — run /validate-idea, /shape-up, or /unpack-problem first
- Very specific: line 30
Create mode is entered on an unvalidated idea at a commitment gate (budget, headcount, or calendar) — escalate to /grill-me first per the …
- Very specific: line 36
Create/edit/validate: artifacts/output/01-discovery/validation-brief.md OR idea-brief.md OR shaped-brief.md
- Distinctive: line 46
Normalise, do not fabricate: Write the prompt to the canonical path and label its provenance:
- Very specific: line 59
Only escalate to /validate-idea, /shape-up, or /unpack-problem when the prompt is genuinely an unvalidated idea at a commitment gate — …
- Very specific: line 68
If entering from shape-up with all assumptions verified, research artifacts are optional.
- Very specific: line 69
If entering from explore-idea, all research artifacts should exist.
- Distinctive: line 84
Follow the step sequence for the selected mode exactly.
- Distinctive: line 100
If unclear, ask: "Are you designing from scratch, refining existing specs, reviewing a design, or turning a prototype into a PRD?"
- Very specific: line 115
If no choice is recorded and prototyping is in scope, pause with a two-track fork: static HTML (recommended for simple exploration) or …
- Distinctive: line 118
Run the deterministic helper with the selected track:
- Very specific: line 272
Create and reverse mode use the same document formats.
- Distinctive: line 294
After PRD + user stories (create step 3 / edit step 3): Write key product decisions:
“Common” is measured: kept by at least 6 in 10 of the 175 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Design
<!-- router-skill -->Bridges exploration and development. Takes validated ideas and turns them into a PRD with requirements, user stories, and detailed product specs with flows, interactions, and visual direction.
When to invoke
- A validated, shaped, or problem-first brief exists at
artifacts/output/01-discovery/and the user wants requirements, user stories, and screen specs produced from it - The user arrives prototype-first — a Figma/Stitch export, a hand-coded HTML mock, or a Vespyr-built prototype — with no PRD yet (reverse mode)
- Specs exist and the user wants them refined (edit mode) or design-reviewed for consistency, accessibility, and gaps (validate mode)
- The user says "design this", "write the PRD", "turn my prototype into requirements", "review the design", or invokes
/design
When NOT to invoke
- No brief and no prototype exist yet — run
/validate-idea,/shape-up, or/unpack-problemfirst - Implementation is the ask and specs already exist and are approved — use
/develop - A PRD exists and only the code is missing, with no design question open
Halt conditions
Any one halts the mode and escalates:
- A prototype is in scope but the
prototype_collaboration.mdcheckpoint is missing or unsigned at handoff scope —prototype_collab_gate.js verifyfails closed; escalate to@product-designer+@tech-lead - The reverse-mode parity gate reports ghost or orphan evidence rows — a requirement that traces to nothing observable is a handoff failure, not a formatting preference
- The Step 0 staleness check reports a stale, broken, or partial freeze — halt loudly; do not proceed on a broken kernel
- Create mode is entered on an unvalidated idea at a commitment gate (budget, headcount, or calendar) — escalate to
/grill-mefirst per the Intent Escalation Ladder
Prerequisites
- Reverse mode (two sub-tracks):
generated— Vespyr-built prototype in the project +artifacts/output/03-strategy/SPEC.mdSpec Kernel (Review 10 §4); norequirements.mdyet.imported— externally-authored prototype (Figma export, Google Stitch, hand-coded HTML) placed inartifacts/input/prototype/; no SPEC.md required — the kernel is reverse-derived (step-0e); norequirements.mdyet.
Create/edit/validate:
artifacts/output/01-discovery/validation-brief.mdORidea-brief.mdORshaped-brief.mdA user prompt is itself a valid brief. This skill's own trigger is the user saying "design this" or "write the PRD", which produces a prompt and not a file at a canonical path. Requiring the artifact above before create mode can start made the primary trigger unable to reach the primary mode: a blind run on a clinical trial brief had to fabricate an
idea-brief.mdand record a deviation, because the only sanctioned route to the missing file was/explore-idea— market and competitive research, irrelevant to an internal, already-funded product.Normalise, do not fabricate. Write the prompt to the canonical path and label its provenance:
# Idea Brief <!-- DERIVED FROM THE USER PROMPT by /design step 01a. Not produced by /validate-idea, /shape-up, or /unpack-problem. No upstream research was commissioned. -->Then mark every unavailable upstream artifact
[ABSENT — not commissioned]rather than running/explore-ideato manufacture one. A brief that says which research you skipped, and why, is more useful than market analysis of a market you are not entering.Only escalate to
/validate-idea,/shape-up, or/unpack-problemwhen the prompt is genuinely an unvalidated idea at a commitment gate — budget, headcount, or calendar. That is the escalation the Intent Ladder requires; it is not an entry fee.artifacts/output/02-research/market-analysis.md(optional if entering viashape-upwith all assumptions verified)artifacts/output/02-research/competitive-analysis.md(optional if entering viashape-up)artifacts/output/02-research/user-personas.md(optional if entering viashape-up)artifacts/output/02-research/journey-map.md(optional — written by/journey-map, consumed by screen specs)artifacts/output/02-research/usability-test-kit.md(optional — ux-researcher kit; results return via ux-research-report.md)
If entering from shape-up with all assumptions verified, research artifacts are optional.
If entering from explore-idea, all research artifacts should exist. If any are missing, load explore-idea first.
Skill chain
- Prev:
explore-ideaorshape-up - Next:
develop
Primary personas
@product-manager— owns requirements, user stories, and PRD scope@product-designer— owns screen specs, design tokens, and product spec
These personas are invoked by step files. The router declares them so the agent knows which reasoning personas are active for this skill before loading steps.
Harness adherence (non-negotiable)
Intent gate: entering Create mode on an unvalidated idea at a commitment gate (budget/headcount/calendar about to be committed)? Escalate to
/grill-mefirst (Intent Escalation Ladder Level 3 —vespyr-dna.md). Ordinary ambiguities stay at Levels 0–2.
- Follow the step sequence for the selected mode exactly. Do NOT skip steps or reorder them.
- Each step file is a contract. Read it fully before executing. Step files override general guidelines.
- Mode detection is automatic. The router decides reverse/create/edit/validate (reverse checked first, A1). Do NOT ask the user which mode unless the intent is genuinely ambiguous.
Mode selection + routing
Detect by checking for existing artifacts + parsing user wording:
Step 0 — Scope Gate (non-skippable): anchor scope first via steps/step-00-scope-and-decision-anchoring.md; the selected mode is the locked track. Step 0 also runs the staleness hook (step0_staleness.js — a stale, broken, or partial baseline hash blocks every mode; script failure halts loudly). step_tracker.js exits 1 if Step 1 begins without a locked track.
Intent triage at gate: Clear (C ≥ 0.85) runs the locked track; Ambiguous (0.50–0.85) halts to a 2–3 Track Fork; Trivial (C < 0.50) executes directly. Gate enforced in all step_tracking modes including off (DNA 4).
- Reverse (checked FIRST — A1) →
prototype-inventory.mdexists ∧ norequirements.md; or the user arrives prototype-first (prototype source present, no requirements). Two sub-tracks resolved at Step 0:generated(Vespyr-built prototype + existing SPEC.md kernel) →steps/step-01d-ingest-prototype.md.imported(external prototype inartifacts/input/prototype/, no SPEC.md) →steps/step-01e-ingest-imported.md, thensteps/step-00e-derive-kernel.md.
- Create → no
requirements.mdorproduct-spec.mdexists (and no prototype-first signal). Route:steps/step-01a-load-prd-brief.md - Edit → artifacts exist, user wants to refine. Route:
steps/step-01b-load-existing.md - Validate → artifacts exist, user wants design review. Route:
steps/step-01c-heuristic-eval.md
If unclear, ask: "Are you designing from scratch, refining existing specs, reviewing a design, or turning a prototype into a PRD?"
Prototype track gate (two supported ways)
The design skill supports two explicit prototype tracks. The user prompt is authoritative; project detection is evidence, not permission to override an explicit request.
| Track | Use when | Runtime | UI contract | Lint contract |
static-html | Early exploration, visual/interaction validation, or a quick prototype | Vanilla HTML/CSS/JS + Tailwind Browser CDN | shadcn-flavored visual language: semantic tokens, familiar component patterns, responsive states; not actual shadcn components | No React, Next.js, or @shadcn/lint requirement. Validate the CDN, token layer, states, and interactions in R1. |
nextjs-component | Engineering-ready prototype, reusable components, or a direct path to implementation | Next.js + React + Tailwind + shadcn source components | Real shadcn components and project components.json | npm run lint must execute @shadcn/lint with zero shadcn/* violations. |
At Step 0, resolve the track in this order:
- An explicit user choice wins (
static HTML/CDNorNext.js/React/shadcn). - A recorded project preference may supply the default, but never overrules an explicit user choice.
- If no choice is recorded and prototyping is in scope, pause with a two-track fork: static HTML (recommended for simple exploration) or Next.js component battery (recommended when the prototype is intended for implementation). Do not silently choose based on
shadcn_gate detect. - Record the selected track and runtime in
project-context.mdthrough the memory controller.
Run the deterministic helper with the selected track:
node .agents/scripts/shadcn_gate.js detect --track <static-html|nextjs-component> --json
node .agents/scripts/shadcn_gate.js verify --track <static-html|nextjs-component> --prototype <file> --jsonFor static-html, verify checks the HTML source, Tailwind CDN reference, token layer, critical states, feedback states, and absence of an unintended React/Next implementation; @shadcn/lint is explicitly not applicable. For nextjs-component, it requires a real component source tree, Next/React/Tailwind/shadcn evidence, and an executed clean npm run lint path with @shadcn/lint. auto remains available for detection-only diagnostics, not for a committed prototype decision.
The shadcn MCP remains optional for the nextjs-component track; MCP never gates design.
Create mode step sequence
- Load PRD Brief →
steps/step-01a-load-prd-brief.md - Define Personas →
steps/step-02a-define-personas.md - User Stories →
steps/step-03a-user-stories.md - Screen States →
steps/step-04a-screen-states.md - Design Tokens →
steps/step-05a-design-tokens.md - Handoff →
steps/step-06a-handoff.md
Prototype collaboration checkpoint
Prototyping is a cross-functional phase, not a designer-only artifact. When a prototype is in scope, @product-designer and @tech-lead must review it before downstream synthesis or handoff. For architecture-impacting work, @architect joins the review.
The checkpoint uses .agents/templates/product/prototype-collaboration-template.md and records prototype-collaboration.md with:
- selected track (
static-htmlornextjs-component) and runtime - prototype URL/file and the observed states/interactions reviewed
- token and responsive mapping
- static-to-component mapping (for
static-html) or component/API boundary (fornextjs-component) - accessibility, feasibility, and implementation risks
- decisions, open questions, and a collaborator state of
SATISFIEDorBLOCKED
A missing checkpoint is a handoff failure. Run node .agents/scripts/prototype_collab_gate.js verify --record <collaboration> --track <track> --prototype <prototype> --scope <handoff|r1-evidence> --json. The r1-evidence scope validates bounded evidence extraction and may preserve an explicit Implementation handoff: BLOCKED; the handoff scope requires accessibility, feasibility, and aggregate sign-off. Stale, mismatched, unsigned, or out-of-scope records fail closed. Do not convert a static prototype into a component implementation implicitly; record the conversion as a separate decision.
Reverse mode step sequence (prototype-first)
Two sub-tracks share the same synthesis spine (Evidence Ledger → PRD synthesis → parity gate). They differ only in ingest source and kernel origin.
generated sub-track (Vespyr-built prototype, existing SPEC.md kernel):
- Ingest Prototype →
steps/step-01d-ingest-prototype.md - Evidence Ledger →
steps/step-02d-evidence-ledger.md - PRD Synthesis (looped create steps 02a→05a, ledger as fixed input) →
steps/step-03d-prd-synthesis.md - Parity Gate →
steps/step-04d-parity-gate.md - Spike Rollback (Sync Contract v2 — invoked only on QA falsification) →
steps/step-05d-sync-contract.md
imported sub-track (external prototype in artifacts/input/prototype/, kernel derived):
- Ingest Imported Prototype →
steps/step-01e-ingest-imported.md - Derive Spec Kernel →
steps/step-00e-derive-kernel.md(reverse-derives SPEC.md; runs only when no kernel exists) - Evidence Ledger →
steps/step-02d-evidence-ledger.md - PRD Synthesis →
steps/step-03d-prd-synthesis.md - Parity Gate →
steps/step-04d-parity-gate.md - Spike Rollback (only on QA falsification) →
steps/step-05d-sync-contract.md
Tracking note. Step 3 above re-runs the create-mode authoring steps 02a→05a with the
Evidence Ledger as fixed input (see step-03d-prd-synthesis.md). Those step files declare
mode: create, so a --skill design-reverse tracker listing shows only the d/e steps and
omits the synthesis work reverse mode actually performs. That is correct — the files are
create steps and their tracking labels are shared — but the tracker's list is not a
complete account of what a reverse run does. Track the synthesis loop against
design-create steps 2–5, and use design-reverse for steps 0e–5d.
What the baseline hash can promise: both sub-tracks record a SHA-256 of the prototype file and compare against it later, so a drift is always detectable. They differ in how strong that promise is. The generated sub-track uses a prototype baseline hash — Vespyr made the file, so it can promise nothing will change it. The imported sub-track uses an import baseline hash — it proves the file you gave me is unchanged since ingest, but cannot detect edits made in the origin design tool (Figma/Stitch). Keeping the imported snapshot in sync with the design tool is a human responsibility; the engine enforces file parity only, never origin parity.
Edit mode step sequence
- Load Existing →
steps/step-01b-load-existing.md - Identify Gaps →
steps/step-02b-identify-gaps.md - Revise →
steps/step-03b-revise.md - Finalize →
steps/step-04b-finalize.md
Validate mode step sequence
- Heuristic Eval →
steps/step-01c-heuristic-eval.md - Consistency Check →
steps/step-02c-consistency-check.md - Accessibility Check →
steps/step-03c-a11y-check.md - Lock & Handoff →
steps/step-04c-lock-handoff.md
Done when
Common to all modes:
@memory-controller session-writehas run, and the chat reply stated the outcome as one plain sentence with the consequence and the next action — never a bare gate token
Create mode:
requirements.mdanduser-stories.mdexist with at least one real story each. A copy ofuser-story-template.mdsatisfies the template's shape and no requirement —template_conformance.jsnow rejects it, and rejects a story that defines no acceptance criterionproduct-spec.md(plusdesign.md) exists with screen states, design tokens, and accessibility coverage- Every screen state table carries the four accessibility states — keyboard, screen-reader, zoom/reflow, reduced-motion — as rows with acceptance criteria, not as a note.
step-04a-screen-states.mddefines them. "Accessibility coverage" is not satisfied by an NFR row and a checkbox: a blind run found create mode carried accessibility in exactly one line of prose, and the dedicated a11y step lives in validate mode's sequence, so the mode where screens are specified never ran it - If the brief names an affected user group, a persona exists for it — a persona, not a paragraph, because a persona is what makes a focus order or an announcement a decision rather than a guess
- The selected prototype track and runtime are recorded in
project-context.md prototype-collaboration.mdexists with aSATISFIEDdisposition whenever prototyping was in scope
Reverse mode:
- The parity gate passes with zero ghost rows, and every FR/NFR, story, AC, screen state, and token decision cites at least one
L-NNEvidence Ledger row
Validate mode:
ux-research-report.mdcarries a Review Gate verdict —VERDICT: [CONFIRMED]/[PARTIAL]/[FALSIFIED], bracketed sotools/eval/lib/ast.jscan read it — preceded by a plain sentence a non-expert can act on- The verdict is derived from the 🟢/🟡/🔴 scorecard that is in the same file, so the number in
Because:matches the findings above it
Edit mode:
gap-map.mdcarries the Review Gate verdict ([CONFIRMED]/[PARTIAL]/[FALSIFIED]), and every revised section has a change record naming the finding it closed- Sections that hit the 2-cycle limit are named as open blocking findings, not carried forward silently
A mode with no completion contract is an unbounded step: the run ends when the agent gets bored, not when the work is done.
Output artifacts
Full paths, so step_tracker.js audit can decide whether this skill ran untracked:
artifacts/output/03-strategy/requirements.mdartifacts/output/03-strategy/user-stories.mdartifacts/output/03-strategy/product-spec.mdartifacts/output/03-strategy/product-spec.htmlartifacts/output/03-strategy/design.mdartifacts/output/03-strategy/gap-map.md— edit mode; holds the Review Gate verdictartifacts/output/02-research/ux-research-report.md— validate mode; also holds its verdict. Required in validate mode, optional in create mode.artifacts/output/03-strategy/measurement-plan.md— optionalartifacts/output/03-strategy/prototype-inventory.md— reverse modeartifacts/output/03-strategy/parity-report.md— reverse modeartifacts/output/03-strategy/prototype-collaboration.md— whenever prototyping is in scope
Also written: requirements.md, user-stories.md, product-spec.md, product-spec.html, design.md, plus measurement-plan.md and ux-research-report.md (optional). When prototyping is in scope, also require prototype-collaboration.md with the selected track/runtime and designer–tech disposition.
Validate mode's verdict lives in ux-research-report.md, not a separate file. That file is required in validate mode — it is the only artifact carrying the gate. It is optional in create mode, where the gate is mechanical (template_conformance.js, anti_generic_check.js, prototype_collab_gate.js) and reports its own pass/fail.
Two surfaces — every mode
/design has four modes and they all end by telling the owner something. The rule is
stated here once because it applies to all of them, not to one step:
- Document surface — the canonical artifact carries the machine record: plain sentence
first, then the bracketed card. Every gate token must be bracketed, because
tools/eval/lib/ast.jsonly reads bracketed forms. - Chat surface — the reply carries the human sentence, the consequence, and the next
action. Never a gate token, and never a card field (
VERDICT:,Survived:,Because:,Next:). A reader who has never seen this engine must be able to act on the sentence alone (.agents/references/vespyr-dna.md§Two surfaces).
State the consequence, not the label. [PARTIAL] is not "partial" — it is "the flow works,
but six states are unusable with a keyboard and none of that is fixed yet; @product-designer
owns it."
Per-step detail lives in the step files; this is the contract they all inherit.
Which gate, when
This skill emits three different gates, and they are not interchangeable. Collapsing them is how a design review ends up reading as a build decision.
| Gate | Emitted by | Vocabulary | Answers |
| Decision | never | [GO] / [RESHAPE] / [NO-GO] | should this be built — owned by /validate-idea and /shape-up, not by /design |
| Review | validate mode (step 04c) | [CONFIRMED] / [PARTIAL] / [FALSIFIED] | are these existing artifacts sound |
| Collaboration | prototype checkpoint | SATISFIED / BLOCKED | did designer and tech agree on the prototype |
So /design must never write [GO] to mean "the design looks fine". Validate mode is judging artifacts that already exist, which is a Review Gate claim (.agents/references/glossary.md).
Canonical document contracts
Create and reverse mode use the same document formats. The templates are authoritative; a compact report is not an acceptable substitute for requirements.md or user-stories.md.
| Artifact | Canonical source | Required output |
requirements.md | .agents/templates/product/prd-template.md | Preserve the numbered PRD structure (§§1–14). Reverse mode fills §14 with the Evidence Ledger and mirrors [OPEN] questions to §12; create mode leaves reverse-only material explicitly not applicable. |
user-stories.md | .agents/templates/product/user-story-template.md | Preserve the backlog tree, Epic → Feature → Story hierarchy, metadata, Business/Technical Requirements, Gherkin AC-H/AC-U/AC-E, UI/UX notes, data notes, out-of-scope items, and open questions. |
product-spec.md / design.md | .agents/templates/product/product-spec-template.md / .agents/templates/product/design.md | Use the existing screen-spec and design-token templates; do not replace them with an evidence summary. |
In reverse mode, every authored persona, FR/NFR, story, acceptance criterion, screen state, and token decision must cite one or more L-NN Evidence Ledger rows. Keep the template's Traces to PRD field and append the ledger IDs to that field (or to the AC evidence line); do not replace the FR reference. A missing template section or required story field is a handoff failure, not a formatting preference.
State & memory integration
At start, run: node .agents/scripts/orchestrator_state.js status, then next.
Neither
statusnornextis read-only. Both createartifacts/output/pipeline-state.json,artifacts/output/sprint-status.yaml, andartifacts/memory/project-context.mdon a tree where they do not exist;nextcan also flip the active phase frompendingtoin-progress. Measured from an empty tree:statusalone creates all three. A blind run was told "statusalone if you want to look without changing anything" and found otherwise — so that sentence was wrong, and it was my own error, not the reader's.To inspect without writing, read the files directly. There is no read-only variant. Also:
@memory-controller load product-manager [design — {feature}]
After PRD + user stories (create step 3 / edit step 3): Write key product decisions:
@memory-controller write active-decisions.md
### [PRODUCT] {feature} design decisions [date: YYYY-MM-DD] [agent: @product-manager]
{key decisions: scope boundaries, prioritized stories, deferred items}
**Status:** activeMemory: Final step in each mode closes with @memory-controller session-write — mandatory per GUARDRAILS §Session Continuity. Include the selected prototype track, runtime, and collaboration disposition when prototyping was in scope.
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.

