user-guide
Skill5 starsChanged 4 months agoMITView on GitHub
Generates a user-facing documentation site from the PRD, user stories, user flows, BRD business rules, and the production application. This skill should be used when the user asks to "generate user documentation", "create the user guide", "build the Mintlify docs site", "generate user-facing docs", "create documentation for users", "run stage 7D", "generate the docs site", or "create getting started guides", or when the transmute-pipeline agent reaches Stage 7D of the pipeline.
In this repository24 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- templates/CLAUDE.mdCLAUDE.md
- audit-a11ySkill
- audit-completenessSkill
- audit-securitySkill
- brdSkill
- docsSkill
- feedbackSkill
- hardenSkill
- implementSkill
- maintainSkill
- optimizeSkill
- polishSkill
- prdSkill
- prelaunchSkill
- redesignSkill
- refactorSkill
- remediateSkill
- scaffoldSkill
- seed-dataSkill
- smokeSkill
- tech-stackSkill
- validate-specsSkill
- verifySkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
42 rules. About 0 are common, 18 are distinctive, and 24 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈18
Distinctive · estimate
This author's own choices. Worth a look.
24
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 17
Read the detailed guide at ${CLAUDE_SKILL_DIR}/references/user-guide-detailed-guide.md for the full agent team architecture, Mintlify …
- Very specific: line 21
Read ./plancasting/_audits/production-smoke/report.md (Stage 7V report).
- Very specific: line 22
Check plancasting/tech-stack.md under the Documentation section.
- Very specific: line 23
Read CLAUDE.md, plancasting/tech-stack.md (product name, design direction, session language, documentation config).
- Very specific: line 24
Check plancasting/tech-stack.md for the Session Language setting.
- Distinctive: line 53
Read 7V report -- confirm PASS or CONDITIONAL PASS.
- Distinctive: line 54
Read Stage 5B report -- identify incomplete features for "Coming Soon" treatment.
- Distinctive: line 55
Read PRD -- extract all user flows (UF-*), screen specs (SC-*), user stories (US-*), feature map, product overview.
- Very specific: line 56
Read Stage 6D help docs (docs/help/, if exists) -- extract adaptable content.
- Very specific: line 57
Read plancasting/tech-stack.md -- extract product name, design direction, session language, documentation platform config.
- Very specific: line 59
Read user flows from plancasting/prd/06-user-flows.md.
- Distinctive: line 68
Create task list for teammates.
- Distinctive: line 87
Verify no PRD/BRD jargon in user-facing content.
- Distinctive: line 88
Verify incomplete features marked "Coming Soon" or omitted.
- Distinctive: line 90
Verify translation parity if multi-language.
- Distinctive: line 95
Verify all teammates completed.
- Very specific: line 96
If PASS or WARN: delete _guide-context.md.
- Very specific: line 111
PASS: All P0/P1 features documented, zero jargon leakage, zero broken links, all docs.json files valid, all navigation entries resolve, …
- Distinctive: line 117
never copy PRD text verbatim -- rewrite for non-technical users.
- Distinctive: line 118
never organize by feature ID -- organize by user journey.
- Very specific: line 119
never use numbered markdown lists for procedures -- always use <Steps>/<Step>.
- Very specific: line 120
never use markdown blockquotes for callouts -- always use <Note>/<Warning>/<Tip>/<Info>/<Check>/<Danger>.
- Very specific: line 122
never reference pages with file extensions in docs.json navigation.
- Very specific: line 123
always validate docs.json as valid JSON.
- Very specific: line 127
always include meaningful description in every page's frontmatter.
- Distinctive: line 128
always cross-reference Stage 5B to avoid documenting incomplete features.
- Very specific: line 129
always extract branding from plancasting/tech-stack.md -- never use placeholders.
- Distinctive: line 130
If multi-language: generate English first, then translate.
- Distinctive: line 131
always include "Next steps" at the bottom of each journey page.
- Distinctive: line 132
never include technical identifiers in user-facing content (no FEAT-*, US-*, SC-*, FR-*, BR-*, UF-*).
- Distinctive: line 133
never reference a screenshot that doesn't exist in the manifest.
- Distinctive: line 134
always wait for page content to fully load before capturing screenshots.
- Very specific: line 135
always include meaningful alt text on screenshot images.
- Very specific: line 137
always use standard markdown image syntax !Alt text for screenshots.
- Very specific: line 138
If screenshot mode is text-only, ALL teammates use <Note> callouts instead of images.
- Very specific: line 139
Use <Accordion> for optional/advanced content that most users can skip.
- Very specific: line 140
Use <Tabs> when the same task has different instructions for different contexts (desktop/mobile, plan tiers).
- Very specific: line 141
If screenshot mode is text-only or images-partial with <50% manifest success, do not mix screenshot and text-only approaches within the …
- Very specific: line 142
always hide UI overlays (Next.js dev badge, cookie consent banners) before capturing screenshots or recordings.
- Very specific: line 143
When an animated recording (MP4) exists for a page, embed it with <video> INSTEAD of the static screenshot -- NOT alongside it.
- Distinctive: line 144
always use PRODUCTION URL for authenticated page captures (screenshots and recordings).
- Distinctive: line 146
always verify captured screenshots/recordings show actual content before embedding.
≈ “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.
Files
User Guide Generation (Mintlify) — Stage 7D
Read the detailed guide at ${CLAUDE_SKILL_DIR}/references/user-guide-detailed-guide.md for the full agent team architecture, Mintlify platform reference, content generation rules, report template, and known failure patterns.
Prerequisites
- Read
./plancasting/_audits/production-smoke/report.md(Stage 7V report). If 7V status is FAIL, stop: "Stage 7D requires 7V PASS or CONDITIONAL PASS. Production must be stable before generating user documentation." If the report does not exist, stop: "Stage 7V has not been run yet. Run Stage 7V first." - Check
plancasting/tech-stack.mdunder the Documentation section. If it says "No -- not needed," stop: "Stage 7D skipped -- user opted out of user-facing documentation in Stage 0." - Read
CLAUDE.md,plancasting/tech-stack.md(product name, design direction, session language, documentation config). - Check
plancasting/tech-stack.mdfor theSession Languagesetting. Generate all content in the session language. If multi-language (session language differs from English), generate English first, then translate.
Inputs
- Production URL: The live application URL
- 7V Report:
./plancasting/_audits/production-smoke/report.md - 5B Report (if exists):
./plancasting/_audits/implementation-completeness/report.md-- identifies incomplete features for "Coming Soon" treatment - 6V Report (optional):
./plancasting/_audits/visual-verification/report.md-- content source for troubleshooting - PRD files:
./plancasting/prd/01-product-overview.md,02-feature-map-and-prioritization.md,03-release-plan.md,04-epics-and-user-stories.md,06-user-flows.md,07-information-architecture.md,08-screen-specifications.md,09-interaction-patterns.md,12-api-specifications.md,18-glossary-and-cross-references.md - BRD files:
./plancasting/brd/11-user-experience-requirements.md,14-business-rules-and-logic.md - Developer Docs (if exists):
./docs/-- adapt (not copy) help content - E2E Constants:
./e2e/constants.ts,./playwright.config.ts
Output
Generate a complete Mintlify documentation site under ./user-guide/. Also generate ./plancasting/_audits/user-guide/report.md.
Multi-language (session language differs from English): Root docs.json with navigation.languages array containing per-language groups + en/ and <session-lang>/ subdirectories. Per-language docs.json files may still be used for navigation overrides.
English-only: Root docs.json with both branding and navigation. Content at root level (no en/ subdirectory).
Target 15-30 pages per language (excluding API reference).
Execution Flow
Phase 1: Lead Analysis and Planning
Perform pre-generation cleanup if ./user-guide/ already exists (check git status/git diff for uncommitted changes first -- stop if found).
- Read 7V report -- confirm PASS or CONDITIONAL PASS.
- Read Stage 5B report -- identify incomplete features for "Coming Soon" treatment.
- Read PRD -- extract all user flows (UF-), screen specs (SC-), user stories (US-*), feature map, product overview.
- Read Stage 6D help docs (
docs/help/, if exists) -- extract adaptable content. - Read
plancasting/tech-stack.md-- extract product name, design direction, session language, documentation platform config. - Derive journey categories from PRD:
- Read user flows from
plancasting/prd/06-user-flows.md. Fall back toplancasting/prd/02-feature-map-and-prioritization.mdif no UF-* entries exist. - Cluster flows by user goal (shared entry point, sequential chain, same domain entity, standalone).
- Name journeys using user-facing language (the user's goal, not feature names).
- Order by natural user progression (onboarding -> core value -> advanced -> admin).
- Map each UF-* to a page within its journey directory.
- Determine concept pages from product overview and glossary.
- Determine public API status (yes/no, OpenAPI spec path).
- Read user flows from
- Build content map -- create
./_guide-context.mdwith product info, journey groups, concept pages, writing style guidelines, branding, language config, incomplete features list, API reference decision, screenshot mode, and feature priority mapping. - Capture screenshots via Playwright MCP browser tools. Generate a screenshot manifest, capture public pages first (no auth required — must succeed), then authenticated pages (best-effort). Hide UI overlays (Next.js dev badge, cookie consent banners) before each capture by injecting CSS via
browser_evaluate. Capture viewport-only screenshots (NOT full-page). Verify each screenshot shows actual content before committing. Save to./user-guide/images/. If the project's.gitignorecontainsscreenshots/, add!user-guide/images/exception. Budget: max 30 screenshots total. Screenshot modes:images(all captured),images-partial(public captured, some auth skipped), ortext-only(only if nothing loads at all). Do not default totext-only— public page screenshots should always succeed regardless of auth status. Tiered failure handling: Tier 1 (public pages) must succeed — failure is a blocker. Tier 2 (authenticated pages) attempt production first; if login fails, try dev server; if dev server shows "Session expired" this is a dev-mode auth race condition — switch back to production. Tier 3 (mode decision) — public success =imagesregardless of auth; some auth skipped =images-partial; onlytext-onlyif both production and dev are completely inaccessible. If./e2e/doc-recordings/exists, optionally capture animated recordings (MP4) for interactive flows — recordings replace static screenshots for the same page. Update_guide-context.mdwith final screenshot mode before spawning teammates. - Create task list for teammates.
Phase 2: Spawn Documentation Teammates
Spawn 3 teammates in parallel for content creation. Teammate 3's validation step waits for Teammates 1 and 2 to complete.
Teammate 1 -- "journey-writer": Write all journey guide pages, introduction.mdx, and quickstart.mdx. For each page, read source UF-, SC-, US-, BR- (or BRL-*) specs. Write using <Steps>/<Step> components. For hero visuals, check the manifest Recordings section first — if an MP4 recording exists, embed with <video autoPlay muted loop playsInline className="w-full aspect-video rounded-xl" src="/images/<name>.mp4"></video> INSTEAD of a static screenshot. If no recording exists, embed screenshots using markdown  syntax — do not use <Frame><img> (causes CDN 403 errors). Add callouts (<Note>, <Warning>, <Tip>, <Info>), and include "Next steps" sections. never start the MDX body with a # Title heading — Mintlify auto-renders the frontmatter title as h1. always escape dollar signs before numbers (\$69, not $69) — MDX interprets $...$ as LaTeX.
Teammate 2 -- "concepts-and-reference-writer": Write concept pages, FAQ (as <AccordionGroup>), troubleshooting (problem -> cause -> solution), changelog (using <Update> components), and API reference (conditional -- only if product has public API without OpenAPI spec). For visual embedding, check recordings first (use <video> tag), then screenshots (use markdown  syntax). always escape dollar signs before numbers. never start the MDX body with a # Title heading.
Teammate 3 -- "config-and-structure": Create root docs.json (branding, theme, colors from plancasting/tech-stack.md) using the validated template from the detailed guide. For multi-language, use navigation.languages pattern in root docs.json. Per-language docs.json may still be used for navigation overrides. Static assets in public/, snippets (prerequisites.mdx, support-cta.mdx). Do not add: colors.anchors, top-level anchors array, top-level languages array, or navigation as a bare array. After Teammates 1 and 2 complete, run validation: mint validate, mint broken-links, mint a11y, jargon scan (two categories: always-flag technical identifiers vs flag-with-context common words), verify  references point to actual files in user-guide/images/, frontmatter description check, translation parity check.
Phase 3: Coordination
Monitor progress, coordinate content consistency, ensure cross-page links are consistent. When Teammates 1 and 2 complete, notify Teammate 3 to proceed with validation.
Phase 4: Review and Audit
- Review Teammate 3's validation results. Run CLI checks if needed.
- Verify no PRD/BRD jargon in user-facing content.
- Verify incomplete features marked "Coming Soon" or omitted.
- Spot-check media references: verify 3-5 random
references in MDX files point to actual files inuser-guide/images/. - Verify translation parity if multi-language.
- Generate
./plancasting/_audits/user-guide/report.mdwith: summary, structure validation, content coverage, screenshot coverage (including animated recordings fields, broken media references, total media size), jargon scan, link integrity, translation parity, CLI check results, and gate decision.
Phase 5: Shutdown
- Verify all teammates completed.
- If PASS or WARN: delete
_guide-context.md. If FAIL: retain it for re-run. - Output summary.
Content Generation Rules
For each guide page, follow this transformation pipeline:
- Identify the user flow from
plancasting/prd/06-user-flows.md-- extract happy path steps. - Enrich with screen specs from
plancasting/prd/08-screen-specifications.md-- use visible labels, not component names. - Add acceptance criteria from
plancasting/prd/04-epics-and-user-stories.md-- transform Given/When/Then into "What you should see" callouts. - Apply business rules from
plancasting/brd/14-business-rules-and-logic.md-- plan-tier restrictions as<Info>, validation rules as<Warning>. - Embed visuals from the manifest in
_guide-context.md-- first check theRecordingssection for an MP4 recording (embed with<video>tag INSTEAD of a screenshot). If no recording, check the screenshot manifest and embed withmarkdown syntax. Place the hero visual after the opening paragraph. If no visual exists, use<Note>callouts. - Rewrite for the audience -- replace ALL technical terms, use product terminology, second person, present tense, imperative mood.
Gate Decision
- PASS: All P0/P1 features documented, zero jargon leakage, zero broken links, all docs.json files valid, all navigation entries resolve, translation parity (if multi-language), CLI checks pass (or N/A if CLI unavailable)
- WARN: P0/P1 fully covered but: P2 coverage gaps, minor style inconsistencies, 1-2 jargon instances in non-critical pages, orphan pages (MDX files not in any navigation), minor
mint a11yissues, or Mintlify CLI checks not tested (CLI unavailable). Deploy with known issues documented. - FAIL: Any of: missing P0/P1 feature coverage, broken navigation (pages in nav but no file), invalid docs.json, >=3 jargon instances, translation structure mismatch, broken media references (image/video references pointing to non-existent files). Fix before deploying.
Critical Rules
- never copy PRD text verbatim -- rewrite for non-technical users.
- never organize by feature ID -- organize by user journey.
- never use numbered markdown lists for procedures -- always use
<Steps>/<Step>. - never use markdown blockquotes for callouts -- always use
<Note>/<Warning>/<Tip>/<Info>/<Check>/<Danger>. - Limit group nesting to ONE level deep in docs.json.
- never reference pages with file extensions in docs.json navigation.
- always validate docs.json as valid JSON.
7a. always include
"theme"in docs.json -- it is a REQUIRED field. Valid values:mint,maple,palm,willow,linden,almond,aspen,luma,sequoia. 7b. always structurenavigationin docs.json as an OBJECT, not an array. For multi-language:{ "languages": [{ "language": "en", "groups": [...] }, ...] }. For single-language:{ "groups": [...] }. A bare array WILL fail. 7c. never usecolors.anchorsor top-levelanchorsin docs.json -- these are deprecated. Use onlycolors.primary,colors.light,colors.dark. - always include meaningful
descriptionin every page's frontmatter. - always cross-reference Stage 5B to avoid documenting incomplete features.
- always extract branding from
plancasting/tech-stack.md-- never use placeholders. - If multi-language: generate English first, then translate. Verify identical structure.
- always include "Next steps" at the bottom of each journey page.
- never include technical identifiers in user-facing content (no FEAT-, US-, SC-, FR-, BR-, UF-).
- never reference a screenshot that doesn't exist in the manifest.
- always wait for page content to fully load before capturing screenshots.
- always include meaningful
alttext on screenshot images. - Maximum 30 screenshots total, maximum 3 per journey page.
- always use standard markdown image syntax
for screenshots. Do not use<Frame><img>wrapping -- it can cause CDN 403 errors on Mintlify's hosting. - If screenshot mode is
text-only, ALL teammates use<Note>callouts instead of images. - Use
<Accordion>for optional/advanced content that most users can skip. - Use
<Tabs>when the same task has different instructions for different contexts (desktop/mobile, plan tiers). - If screenshot mode is
text-onlyorimages-partialwith <50% manifest success, do not mix screenshot and text-only approaches within the same guide -- use<Note>callouts consistently. - always hide UI overlays (Next.js dev badge, cookie consent banners) before capturing screenshots or recordings.
- When an animated recording (MP4) exists for a page, embed it with
<video>INSTEAD of the static screenshot -- NOT alongside it. The video replaces the screenshot. - always use PRODUCTION URL for authenticated page captures (screenshots and recordings). Dev servers have known race conditions with real-time auth providers where the initial token fetch gets aborted during navigation, causing permanent "Session expired" error boundaries.
- For animated recordings, always trim the login/navigation portion from auth recordings during conversion (use
ffmpeg -ss <seconds>). The video should start on the feature page with content visible. - always verify captured screenshots/recordings show actual content before embedding. Check for: loading spinners, "Session expired" screens, error boundaries, blank pages, cookie banners, dev badges. Retry with longer wait times or overlay hiding if needed.
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.

