JoshuaShepherd
- 405 skills
- 0 followers
- 1 repo stars
- 2 weeks ago last updated
- ▌ Typography Polish 3 · joshuashepherdAudit typography for reading comfort, heading hierarchy, responsive scaling, and content rendering surfaces.
- ▌ Supabase Security Audit 2 · joshuashepherdAudit Supabase security — RLS policies, exposed secrets, API route auth, storage security, and role grants.
- ▌ Pathway Content Author · joshuashepherd bundleGenerate the full markdown content for any Movemental pathway page — all twelve canonical sections — in the documented voice and style of a given author, grounded in that author's own books and research in the repo (the books corpus and movement_leader_research / dashboard-content). GENERAL skill: works for any tenant author and any pathway theme once an author and a theme are given. Use it whenever someone wants to write, draft, or generate the content for a pathway page — e.g. "write the [theme] pathway in [author]'s voice", "generate the pathway content for [author] on [topic]", "draft the [slug] pathway from [author]'s books", or loosely "do the [author] pathway on [theme]". Produces corpus-grounded, voice-applied, four-necessities-complete pathway markdown (the Phase 1 narrative content) that feeds the code-level pathway-author and pathway-audit skills. It is the CONTENT layer; it does not build the page template or the code-level [slug].ts.
- ▌ Analytics Audit · joshuashepherdAudit the analytics implementation across all providers (GA4, PostHog, Vercel Analytics, Supabase) — verifies events fire correctly, tenant_id is set on all events, no duplicates, user identity is wired, and privacy/GDPR compliance is met.
- ▌ Analytics Setup · joshuashepherdBootstrap the complete analytics stack for this Next.js 15 + Supabase project — GA4 (org-wide + per-tenant), PostHog, Vercel Analytics, and Supabase-native analytics. Runs sub-skills in sequence and validates each layer.
- ▌ Stitch Intake Audit · joshuashepherdPhase B/01 of the Movemental tenant migration — fetch and cache Google Stitch wireframe screens locally, map each screen to its Next.js route, inventory which routes are still PlaceholderPage scaffolds, and produce a section-level gap audit before any React conversion. Use when starting a stitch migration, when the user says "intake the stitch screens", "cache the wireframes", "gap audit", "screen to route map", or before running stitch-page-port. Writes docs/build/notes/stitch-screen-route-map.md and stitch-gap-audit-<date>.md. Tenant-agnostic: reads TENANT_MANIFEST.md for the Stitch project, prompt library, and route scope. Does NOT convert to React (that's stitch-page-port) and does NOT change tokens (that's stitch-token-bridge).
- ▌ Stitch Token Bridge · joshuashepherdPhase B/02 of the Movemental tenant migration — map the Google Stitch black-and-white wireframe palette (the layout-proof grayscale) onto the tenant's production semantic design tokens, plus the typography and spacing bridge. Run ONCE after stitch-intake-audit and before any page conversion. Use when the user says "token bridge", "map the wireframe colors", "set up the design tokens for the stitch migration", or before stitch-page-port. Writes docs/build/notes/stitch-token-bridge.md and makes minimal, approved edits to globals.css / tailwind.config.ts. Tenant-agnostic: the production theme (e.g. Editorial Indigo / Atelier Amber) comes from TENANT_MANIFEST {{DESIGN_THEME}}. Prevents raw #111/#666 hex and wireframe fonts from leaking into production components.
- ▌ Movement Leader Harvest Ingest · joshuashepherdIngest harvest-manifest.json items into Supabase content tables for a movement leader tenant — archive_items or content_items for FETCH articles, videos and podcast_episodes for LINK media, draft-by-default with namespaced slugs. Uses leaders.manifest.json for org/author IDs. Run after content-harvest (and media-transcribe for A/V). Distinct from movemental-leader-corpus-upload (dossier JSONB) and demo-pack ingest (synthetic seed).
- ▌ Movement Leader Platform Build · joshuashepherd bundleAudits a movement-leader tenant against the master component checklist, writes a gap-remediation build prompt to docs/build/prompts/, then implements missing pages and components in the repo's existing design system with full type-safety compliance. Use when running the platform checklist, auditing component gaps, building remaining pages, platform build, master-component-checklist, or remediating missing shell/themes/ library/formation routes. Run from any movement-leader repo root.
- ▌ Movemental Committed Voice Bio · joshuashepherdGenerate the Committed Voices onboarding entry for a specific movement leader from their research dossier. Use when asked to "draft the committed-voice bio for <leader>", "generate Committed Voices entry for <slug>", or "extract committed-voices markdown for <author>". Reads the leader's research dossier under movemental-ai/docs/movement_leader_research/<slug>/ and emits one structured markdown file (YAML frontmatter + slightly-warm third-person editorial bio) that maps directly onto the CommittedVoice type rendered at /onboarding/committed-voices.
- ▌ Pathway Faq · joshuashepherdGenerate FAQ and QA content for an Alan Hirsch pathway page. Produces 6–10 GEO-optimized Q&A pairs grounded in Alan's books, written in his voice, covering common confusions, objections, term clarifications, and distortion warnings. High SEO/GEO value — AI engines cite FAQ pages directly.
- ▌ Deploy To Vercel · joshuashepherd bundleDeploy applications and websites to Vercel. Use when the user requests deployment actions like "deploy my app", "deploy and give me the link", "push this live", or "create a preview deployment". Supports both interactive login and token-based authentication.
- ▌ E2e Studio Tests · joshuashepherdRun e2e tests in the Studio app. Use when asked to run e2e tests, run studio tests, playwright tests, or test the feature.
- ▌ Supabase Fix Rls · joshuashepherdTake an RLS finding from supabase-security-audit (or a free-form description of an RLS issue) and apply a reviewed fix via the Supabase MCP. Handles the four common patterns — RLS disabled, RLS enabled with no policies, overly permissive policies, missing tenant scope — plus a bonus path for excessive role grants. Every write is gated on explicit operator approval. Use when asked to "fix RLS on X", "lock down table Y", "harden the audit findings", or "remediate Supabase advisories".
- ▌ Tenant Migration · joshuashepherdEnd-to-end migration skill for spinning up a new movement leader platform from an existing tenant baseline. Covers repo setup, Vercel deployment, env vars, org/tenant config, Supabase storage, Claude/Cursor project files, docs cleanup, front-end customization, and post-migration verification. The default migration strategy for all new tenant platforms.
- ▌ Vercel Analytics · joshuashepherdSet up Vercel Analytics (page views, custom events) and Speed Insights (Core Web Vitals) for this Next.js 15 project. Zero-config on Vercel deployments — no API key needed for basic tracking.
- ▌ Workspace Author · joshuashepherdWrite new workspace documents — articles, research notes, project docs, and ebook sections — following Brad Brisco's voice identity, EEAT theme taxonomy, and the workspace frontmatter conventions. Use when creating new content for docs/workspace/.
- ▌ Movement Leader Content Harvest · joshuashepherd bundleComprehensively discover and migrate every external content source for one movement leader except books — articles, blogs, YouTube, Vimeo, podcasts, audio, sermons, talks, interviews, journals, academic papers, courses, newsletters, and long-form social. Runs the full 00–09 content-gathering pipeline for any leader slug in movement-leader-websites. Use when harvesting content, migrating external sources, running content gathering, or closing migration gaps for a movement leader writer.
- ▌ Movemental Leader Corpus Upload · joshuashepherd bundleLoad a movement leader's research into Supabase movement_leader_corpus_data and the related onboarding tables (corpus_review_items, onboarding_tasks, movement_leader_welcome_letters, organizations.onboarding_state). Runs the research:bridge → etl → publish → publish-letter → verify pipeline that compiles the on-disk research tree (profile/, content/, network/, reflected-understanding) into the typed JSONB columns (identity, biography, theology, voice_analysis, calling_profile, books, articles, audio, videos, frameworks, organizations, reflected_understanding_md). Use this WHENEVER asked to push/load/sync/ETL a movement leader's research or dossier into Supabase, wire up movement_leader_corpus_data, fill empty dossier dimensions (videos, audio, frameworks, "A Letter"/reflected understanding, theology, calling), re-run the corpus pipeline, or onboard a leader's corpus — even if the request says "dossier" or "/profile" rather than "corpus_data". Knows the file→column→parser map and the critical substrate rule (s
- ▌ Supabase Postgres Best Practices · joshuashepherd bundlePostgres performance optimization and best practices from Supabase. Use this skill when writing, reviewing, or optimizing Postgres queries, schema designs, or database configurations.
- ▌ Translation Audit · joshuashepherdDeep linguistic audit of translated books in the Alan Hirsch corpus. Randomly samples ~10,000 words from a target language (Brazilian Portuguese, Latin American Spanish, German, European Portuguese, or Mandarin Chinese) alongside the corresponding English source, then grades the translation on a 100-point rubric across six linguistic dimensions. Produces a full markdown report with specific issues, exact locations, improvement recommendations, and an agentic confidence overview for non-speakers. Use when asked to audit, grade, review, or quality-check any translation in the corpus. Trigger phrases: "audit [language] translation", "grade the Spanish", "check translation quality", "review the German corpus", "translation QA".
- ▌ Dashboard Agent Audit · joshuashepherdSign into the Movemental Studio dashboard via Playwright as EACH movement-leader tenant (or a chosen subset), traverse a route span (dashboard, content-books, settings, profile), surface console/API/runtime errors per tenant, and write per-leader JSON audit reports. Use when asked to dashboard audit, run span audit, surface dashboard errors, check whether each tenant/leader is good to go, Playwright audit, traverse dashboard routes, agent audit dashboard, or smoke-test the dashboard as signed-in users. Works headless on Ubuntu/WSL/CI — do not default to Chrome DevTools MCP.
- ▌ Poll Opinion Research · joshuashepherdResearch public opinion, polling data, and sentiment trends using Pew Research, Gallup, Ipsos, Eurobarometer, YouGov, and other major polling organizations. Produces a structured report with data tables, trend analysis, and full source attribution.
- ▌ Tenant Backend Parity · joshuashepherdPhase C/11–16 of the Movemental tenant migration — bring a tenant's Supabase-backed data layer to parity with the reference repo so stitch-migrated UI can wire real data. Walks: 11 Supabase org & tenant identity audit (MCP), 12 six-layer type-safety chain port (Drizzle → Zod → Services → Routes → Hooks → UI) + validate:all, 13 services & API routes parity, 14 hooks layer parity, 15 tenant.config.ts ↔ DB feature alignment, 16 backend validation & cutover. Use when the user says "port the backend", "type safety chain", "services/hooks parity", "align the tenant config with the database", "feature flags vs db", "verify tenant org", or finishes the stitch UI and needs live data. Tenant-agnostic: org id, project id, reference repo, and leak patterns come from TENANT_MANIFEST.md. Delegates to tenant-migrate, type-safety-chain, validate, tenant-check.
- ▌ Integrate Stitch Template · joshuashepherdFully integrate a Google Stitch screen (or set of screens) into the movement-leader-html-templates platform. Given Stitch instructions (project title + ID and one or more screens with titles + screen IDs), this scaffolds the template folder, pulls the HTML + screenshot from Stitch, gives it a Framer-style name, copy-edits the content to fit the movement-leader author/pastor platform without breaking the UI, swaps placeholder/external images for local repo assets, extracts design tokens, and rebuilds + validates the catalog. Runs in both Claude Code and Cursor. Use when adding a new template from a Stitch design.
- ▌ Movement Leader Harvest Manifest · joshuashepherd bundleBuild or reconcile harvest-manifest.json from an existing movement leader content set — docs/harvest markdown, legacy docs/articles|videos, CSV/JSON catalogs, and LINK-only media rows. Classifies type/disposition, assigns stable ids, validates file paths for ZenWrite bundle upload and pnpm harvest:ingest. Use when creating a manifest, fixing manifest drift, preparing a harvest zip, or mapping a folder of files to ingest-ready rows for any leader slug.
- ▌ Movement Leader Media Transcribe · joshuashepherdTranscribe YouTube and podcast items from a movement leader's harvest-manifest.json — yt-dlp auto-captions for video, Whisper for podcast RSS/mp3 URLs. Writes verbatim transcripts to docs/harvest/transcripts/ and updates the manifest. Run after movement-leader-content-harvest and before movement-leader-harvest-ingest when LINK/TRANSCRIBE media needs text.
- ▌ Supabase Analytics · joshuashepherdSet up Supabase-native analytics for this multi-tenant platform — custom analytics_events table (RLS-scoped per org), per-tenant views, pg_stat_statements for query performance, and Logflare for log analytics.
- ▌ Workspace Organize · joshuashepherdAudit and fix workspace doc organization — frontmatter hygiene, section placement, sidebar rendering, and frontend-safe content. Ensures docs/workspace/ files render cleanly with no metadata leaking into the reader view. Use when adding docs, reorganizing sections, or before deploying.
- ▌ Workspace Strategy · joshuashepherdStrategic planning for the workspace doc system — content roadmaps, agentic workflow design, context coding architecture, research priorities, and reflection. Use when deciding what to build, write, or research next in docs/workspace/.
- ▌ Oatmeal Template Audit · joshuashepherdAudit and refine an HTML page in `1-html/by-template-family/ml-templates/oatmeal/` against the canonical Oatmeal template fingerprint — tokens, primitives, components, layout vocabulary, decorative motifs. Use whenever a new page is built in the oatmeal directory, when a page is being refined for template fidelity, or when the user says "make this look like the oatmeal template," "audit this against oatmeal," "is this on-brand for oatmeal," or "apply oatmeal styling." Complements `movemental-page-auditor` (which judges narrative/proof) and `page-audit` (which judges DESIGN.md compliance) — this skill judges **template fidelity only**.
- ▌ Tenant Structural Port · joshuashepherdPhase D (optional deep port) of the Movemental tenant migration — port STRUCTURAL code parity from the reference repo (e.g. alan-hirsch) directly, rather than from Stitch wireframes, when the design-first stitch path is insufficient. Covers: design-chain audit so reference code lands with TENANT design tokens not the reference's, course infrastructure diff & port (hooks, custom APIs, learn player, certificates, cohort discussions, availability), course frontend parity verification, page inventory & route normalization, page-structure port with tenant design, tenant content/copy substitution, and final validation/cutover. Use when stitch-migration-validate flags hook/data/UX gaps, or the user says "port the course player from the reference", "structural parity", "match alan-hirsch course UX", "normalize the routes", "no source-tenant leaks". Run AFTER the stitch + backend phases. Tenant-agnostic via TENANT_MANIFEST.md.
- ▌ Movemental Welcome Letter Publish · joshuashepherdPublish a finished Movemental dashboard welcome letter to the production Supabase database so the named movement leader sees it on their next dashboard load. Use whenever the user says "publish the welcome letter for <author>", "push <author>'s welcome letter live", "ship the dashboard letter", "publish welcome letter from <file>", or any phrasing that combines a finished welcome letter with making it visible to the leader. Companion to the movemental-welcome-letter skill, which generates the draft; this skill takes the finished draft (file or pasted text) and wires it into the right tenant. Run this after the human has read the draft and is ready to make it live.
- ▌ Pathway Quotes · joshuashepherdFind, verify, and curate quotes from Alan Hirsch's books for pathway pages. Conducts corpus research across all 13 books, extracts verbatim passages, verifies attribution, and produces 3–5 citable pull quotes per pathway — definitional, prophetic, and pastoral. These are the sentences AI engines will cite verbatim.
- ▌ Analytics Dashboard · joshuashepherdBuild React analytics dashboard components for this platform — KPI metric cards, time series charts (Recharts), funnel views, and per-tenant data scoping. Uses shadcn/ui + Recharts, wired to React Query hooks over the analytics_events Supabase table.
- ▌ Migrations Workflow · joshuashepherdManage the full Drizzle ORM migration lifecycle — generate, review, apply, rollback, and sync with Supabase. Covers the safe workflow for schema changes, CI migration checks, and multi-environment migration strategy. Use when making any database schema change.
- ▌ Telemetry Standards · joshuashepherdPostHog event tracking standards for Supabase Studio. Use when reviewing PRs for telemetry compliance or implementing new event tracking. Covers event naming, property conventions, approved patterns, and implementation guide.
- ▌ Vercel Deploy Audit · joshuashepherdAudit a Vite+Express workspace for Vercel deployment issues — API connectivity, env vars, serverless bundling, CORS, symlinks, rewrites, and build config. Returns prioritized findings with fixes.
- ▌ Movemental Page Auditor · joshuashepherd bundleAudit any Movemental page (HTML mockup or production React) across six passes — narrative sequencing, copy & language, typography/hierarchy, UI/proof/demonstration, proof burden, and cross-site role — and return a structured, ranked critique. Use this skill whenever the user asks to audit, review, critique, diagnose, or "be tough on" a Movemental page (`/`, `/fragmentation`, audience pages, articles, `/book`, `/system`, `/walkthrough`, FAQ, `/assess`, `/assessment-new`, audience concept pages), or asks questions like "is this page working?", "what's weak here?", "what's missing?", "should this section exist?", "does this page duplicate /X?". Judges against Movemental canon (two intelligences, the AI Stewardship Sequence, six stages, formation, inspectability, field guide), not generic design best practices. Use this skill even when the user doesn't explicitly say "audit" — if they are asking for a serious, opinionated read of a whole page (not just one section), this is the right tool. Complements, does not r
- ▌ Movemental Publish Gate · joshuashepherd bundlePre-publication quality gate for research and authoritative content headed to movemental.ai. Run before posting any article, research synthesis, field guide, footnote source, /voices or /footnotes claim, or long-form doc. Checks ten red-light gates — AI-slop and register-jargon, citation and evidence integrity, factual/hallucination risk, doctrine drift (movement leaders vs orgs, "Trusted voices"), EEAT self-evidence, mechanical correctness, and legal/sensitivity — and returns a RED / YELLOW / GREEN verdict with the exact blocking items and which fix skill to run. A slop, mistake, and quality-control checker that decides "is this safe to publish yet?" Use whenever the user says publish / post / ship / "is this ready" / "check this before it goes live" / "QC this" about Movemental content. Routes to plain-prose, movemental-prose, and movemental-narrative-audit rather than duplicating them.
- ▌ Platform Demo Architect · joshuashepherdInteractive 9-step process to design a cutting-edge, best-case platform demo for a specific audience — audience intel, problem thesis, capability map, personality/theming, then auto-generated sitemap, component specs, and agent roster (Next/Vite, Tailwind, Supabase RLS, Vercel, Stripe+Shopify, Resend, tenant-scoped agents). Opinionated, high-signal; use when scoping a tailored tenant platform or stakeholder demo.
- ▌ Pathway Content · joshuashepherdMaster skill for authoring any content type on an Alan Hirsch pathway page. Understands the 12-section architecture, all five pathways, book-to-pathway mapping, Alan's voice system, and where every piece of existing content lives. Use this when writing or auditing any pathway page content — or delegate to a specialized sub-skill (pathway-faq, pathway-quotes, pathway-scripture, etc.) for a single section.
- ▌ Pathway Profile · joshuashepherdResearch and write profiles of people and stories for Alan Hirsch pathway pages. Produces leader profiles, movement founder stories, and biographical narratives drawn from Alan's books and research — people who embody the pathway's principles. Use for the Content and Case Study sections, or for standalone biographical content.
- ▌ Liveblocks Workspace · joshuashepherdWork with the Liveblocks + BlockNote collaborative workspace — add features, debug content rendering, configure rooms, manage auth, set up webhooks, add comments/threads, implement version history, or troubleshoot collaborative editing. Use when touching any workspace/docs collaboration code.
- ▌ Nextjs Supabase Auth · joshuashepherdExpert integration of Supabase Auth with Next.js App Router Use when: supabase auth next, authentication next.js, login supabase, auth middleware, protected route.
- ▌ AI Lab Notebook Gemini · joshuashepherdExpert guidance for managing and optimizing the AI Lab Notebook's Gemini integration in this repo — context cache vs File API decisions, corpus manifest lifecycle (48h expiry), Gemini/Claude backend switching, model↔cache ID locking, grounded source discovery, and build-time upload pipeline. Use whenever the user mentions the ai-lab notebook, the Gemini corpus, GEMINI_CORPUS_CACHE_ID, GOOGLE_GENERATIVE_AI_API_KEY, notebook chat failures, "corpus not loading", manifest expired, Google Search grounding in sources discover, or scripts/gemini-corpus-upload.ts / scripts/gemini-corpus-cache.ts. For generic Gemini API questions unrelated to this notebook, defer to the `gemini-api` skill.
- ▌ Pathway Glossary · joshuashepherdGenerate glossary terms for an Alan Hirsch pathway page. Produces 5–10 movemental vocabulary definitions drawn from Alan's books, using his language, linked to the full glossary. These are the terms a reader needs to inhabit this pathway — not a dictionary, but a language school.
- ▌ Pathway Overview · joshuashepherdWrite or rewrite the Overview section for an Alan Hirsch pathway page — the 400–600 word thematic summary that serves as the invitation to enter. Combines corpus research, Alan's voice, and the pathway's core reframe question. This is the first content users read below the hero and core question — it must be the best prose on the page.
- ▌ Stitch Migration Validate · joshuashepherdPhase B/10 of the Movemental tenant migration — the UI validation gate run after a stitch-page-port slice (or after Prompts 03–09) and before merge. Runs build/lint/type checks, greps for the forbidden patterns (PlaceholderPage, dangerouslySetInnerHTML, raw wireframe hex), does a Stitch-PNG-vs-rendered visual spot check, updates COMPONENT_CHECKLIST.md and REFERENCE_PAGE_COMPARISON.md, and emits a conversion report. Use when the user says "validate the stitch slice", "run the UI gate", "check the migration before merge", "is this slice done", or finishes a page-port. Tenant-agnostic: leak grep patterns and route scope come from TENANT_MANIFEST.md. This is the UI gate; full backend validation lives in tenant-backend-parity.
- ▌ Tenant Migration Playbook · joshuashepherd bundleOrchestrate the end-to-end migration of a Movemental thought-leader tenant — Google Stitch B&W wireframes → production React (L4/L5) → Supabase type-safety backend parity → tenant config alignment → validation. Use when onboarding or migrating any leader site under movement-leader-websites/ (e.g. michael-cooper, brad-brisco), when the user says "run the tenant migration", "migrate the stitch wireframes", "stitch playbook", "tenant onboarding pipeline", or references TENANT_MANIFEST.md / COMPONENT_CHECKLIST.md / MASTER_PLAYBOOK.md. This is the hub: it reads the manifest, scopes the slice, and dispatches to stitch-intake-audit, stitch-token-bridge, stitch-page-port, stitch-migration-validate, tenant-backend-parity, and tenant-structural-port. Tenant-agnostic — all tenant values come from the manifest, never hardcoded.
- ▌ Pathway Scripture · joshuashepherdWrite exegetical Scripture content for Alan Hirsch pathway pages. Produces a primary passage with theological commentary, 3–5 supporting references, and a redemptive history thread — all in Alan's voice. This is theology, not homiletics. Scripture is woven into the argument, grounding the pathway's concepts in biblical reality. Use when writing the Scripture section of any pathway page.
- ▌ Ml Template From Reference · joshuashepherd bundleBuild an HTML/CSS/JS movement-leader template (home + content library + articles) whose visual design matches a provided reference image (or images) exactly. The movemental (alan-hirsch) navigation structure is fixed and consistent across leaders; everything else — color, type, layout, density, motifs — is extracted from the reference. The reference image(s) and an extracted style-spec are archived inside the template directory so future passes can re-consult them.
- ▌ Movemental Narrative Audit · joshuashepherd bundleAudits any platform page’s copy, narrative arc, mental models, and business positioning against Movemental’s canonical story (dual intelligences, six-stage progression, infrastructure thesis, scenius/credibility). Use when the user asks for narrative alignment, messaging audit, “does this match our model”, framework drift, editorial re-alignment, or copy QA against the fragmentation book, fragmentation page, home page, docs/articles, or docs/design/DESIGN.md.
- ▌ Nonprofit Pricing Research · joshuashepherd bundleProduce transparent, defensible pricing intelligence for AI-literacy and applied AI training engagements for nonprofits. Runs a scoping intake, gathers real market comps, and outputs a client-ready memo plus internal pricing workbook with low/base/high ranges and a sensitivity table. Use when asked to price an AI training engagement, write a nonprofit AI training SOW, research what similar programs cost, or prepare a proposal for an AI rollout. NOT for legal, tax, or HR compensation advice.
- ▌ Pathway Case Study · joshuashepherdResearch and write case studies for Alan Hirsch pathway pages. Produces 2–4 case studies per pathway — biblical, historical, and contemporary — with card-level summaries for the pathway page and full narrative pages for dedicated URLs. Draws from Alan's books, documented movements, and his research. Each case study follows the pathway's own arc (U-shape for Metanoia, reframes for Reframation, etc.).
- ▌ Postgres Best Practices · joshuashepherd bundleComprehensive PostgreSQL reference covering performance optimization, schema design, indexing, query patterns, connection management, security, and advanced features. Use this skill when writing, reviewing, or optimizing Postgres queries, schema designs, migrations, or database configurations.
- ▌ Supabase Security Audit · joshuashepherdAudit a Supabase project for security issues — RLS, auth config, role grants, storage, functions, and exposed secrets. Returns prioritized findings with fix recommendations.
- ▌ Movemental Tenant Provision · joshuashepherdProvision Movemental movement-leader tenants end-to-end on Supabase — runs Phase 0 prerequisite security/auth gates first, then per-leader identity → auth user → user_profile → owned org → owner membership. Targets the canonical author cohort (Alan Hirsch, Brad Brisco, Liz Rios, Rowland Smith, JR Woodward, Lucas Pulley, Jamie Roach) and the Movemental admin org. Use when asked to "wire up Movemental tenants", "onboard a movement leader", "provision the author orgs", or "run the tenant-wiring prompt".
- ▌ Pathway Article Gen · joshuashepherdGenerate standalone articles for the Alan Hirsch platform — grounded in pathway themes, written in Alan's voice, with full corpus research, SEO/GEO architecture, and proper citations. Produces Tier 1 pillar articles (3,500–4,500 words), Tier 2 cluster articles (2,200–3,200 words), and Tier 3 long-tail articles (1,500–2,200 words). Each article maps to a pathway pillar and connects to the broader content ecosystem.
- ▌ Supabase Add Tenant User · joshuashepherdProvision a new multi-tenant user on a Supabase-backed app with one auth user → one user_profiles row → one owned organization → one organization_memberships row. Tuned for the Movemental / alan-hirsch visual-editor schema (public.user_profiles, public.organizations, public.organization_memberships). Use whenever the user asks to "add a new user", "invite a creator", "onboard a tenant", or "add a member to an org".
- ▌ Pathway Visualization · joshuashepherdGenerate visualization design briefs for Alan Hirsch pathway pages. Produces detailed specifications for diagrams, charts, and visual models that illustrate each pathway's core framework — suitable for handoff to a designer or asset-generate tool. Understands Alan's existing visual models (U-Shaped curve, mDNA hexagon, APEST pentagon, etc.) and translates them into actionable design briefs.
- ▌ Movemental Site Template Publish · joshuashepherd bundlePublish a Stitch-authored HTML/Tailwind one-page site mockup into the Movemental Studio front-end-template catalog so it shows up as a selectable option in the onboarding template selector. Use WHENEVER the operator hands over a Stitch (or any standalone Tailwind) HTML file and wants it "uploaded to Supabase", "added as a template", "added to the template selector / options", "published as a front-end template", or "turned into a site template". Takes the raw HTML, derives catalog metadata, saves the reference HTML, screenshots a desktop + mobile preview, uploads the previews to the `template-previews` Storage bucket, upserts a `front_end_templates` row (status=published), and verifies the template now appears in the published catalog the selector reads. Runs in the `movemental-visual-editor-main` repo.
- ▌ Transformative Learning Collaborator · joshuashepherdExpert peer collaborator for transformative learning design — thinks with you about course architecture, content alignment, pedagogical integrity, and how Alan Hirsch's vision translates into formation experiences that actually change people. Use when you need a thinking partner on course design, content strategy, learning outcomes, or alignment between Alan's frameworks and the course experience.
- ▌ Javascript Typescript Typescript Scaffold · joshuashepherdYou are a TypeScript project architecture expert specializing in scaffolding production-ready Node.js and frontend applications. Generate complete project structures with modern tooling (pnpm, Vite, N
- ▌ Route Audit · joshuashepherd bundleWalk, verify, repair, and sign off Movemental Studio / ZenWrite UI routes using Chrome DevTools MCP (diagnose) and Playwright specs (certify). Use for any portion of the app: a single view (?view=create, Money, Home), a phase (1-public, 3-studio-content), public reader URLs (/articles/:slug, /pricing), or a full pre-launch sweep. Triggers include: route audit, walk the app, check every page, are the routes working, audit Create/Organize/Reach/Money, sign off routes, resume route audit, verify routes.manifest, run audit:routes, or "is /?view=X broken?". Reads routes.manifest.yaml + docs/audit/routes/STATE.md; never marks signed_off on the same turn as a fix.
- ▌ Agent Context · joshuashepherdBuild or modify user context systems — profile assembly, RunContract payloads, context schemas, catalog validation. Use when working on how user data flows into agents.
- ▌ Agent Handoff · joshuashepherdConfigure multi-agent handoffs — agent registry, delegation rules, and routing logic. Use when setting up agent-to-agent transfers.
- ▌ Build Context · joshuashepherdWork on agent context — use when asked to build, modify, or debug the user context system for an agent, including context schemas, RunContract mapping, context block rendering, or catalog slug validation.
- ▌ Asset Animate · joshuashepherdCreate micro-animations from still images — parallax, Ken Burns, floating particles, breathing effects, cinemagraphs. Generates CSS animations, GSAP code, or short video/GIF from a static image. Use for hero backgrounds, social posts, and subtle motion design.
- ▌ Asset Deliver · joshuashepherdPackage generated prompts or assets into a polished interactive HTML page with tabbed navigation, one-click copy buttons, and visual polish. Use to deliver prompt sets, asset galleries, or comparison views as a premium shareable artifact.
- ▌ Video Consult · joshuashepherdConsult on video production for a course — scripture grounding, narrative arc, storyboard, scene structure, and script. Bridges Alan's voice, course theology, and Remotion video production. Use when planning or authoring video content for any course.
- ▌ Build Prompt · joshuashepherd bundleTurn a rough idea, voice memo, or half-formed brief into a fully engineered markdown prompt at docs/build/prompts/<slug>.md for any agentic coding situation (Claude Code, Cursor, Codex, Aider, etc.). Embeds the four prompt-engineering pillars — Clear Communication, Context, Definition of Done, Iteration — plus role, output format, guardrails, and a closed-loop acceptance protocol. Interprets confidently; asks only when truly ambiguous; iterates with the user until the Definition of Done is met.
- ▌ Book Convert · joshuashepherdPhase 01 of the books pipeline — convert any PDF or EPUB book into lossless, paginated, footnoted Markdown using the right tool for the input type (Marker for native-text PDFs, MinerU for scanned PDFs, Pandoc for EPUB, Mistral OCR as hosted fallback). Auto-routes by file type, writes per-chapter MD with `<!-- page: N -->` markers and `[^N]` footnotes, and lands the output in `corpus/alan_hirsch/<book-slug>/chapters/`. Use when ingesting a new book source file from `_inbox/` into the corpus. Trigger phrases: "convert this PDF", "ingest this book", "run phase 01", "PDF to markdown".
- ▌ Course Audit · joshuashepherdAudit the course learn experience — checks section components, sidebar rendering, progress tracking, responsiveness, design tokens, and accessibility. Use before shipping a course.
- ▌ Ckm UI Styling · joshuashepherd bundleCreate beautiful, accessible user interfaces with shadcn/ui components (built on Radix UI + Tailwind), Tailwind CSS utility-first styling, and canvas-based visual designs. Use when building user interfaces, implementing design systems, creating responsive layouts, adding accessible components (dialogs, dropdowns, forms, tables), customizing themes and colors, implementing dark mode, generating visual designs and posters, or establishing consistent styling patterns across applications.
- ▌ Tam Audit · joshuashepherdAudit the TAM master list for gaps, staleness, scoring inconsistencies, demographic imbalances, and missing research — produces an actionable health report
- ▌ Tam Score · joshuashepherdScore and rank movement leader candidates using the Seven Gates and 100-point Movemental rubric
- ▌ Stitch Design · joshuashepherd bundleUnified entry point for Stitch design work. Handles prompt enhancement (UI/UX keywords, atmosphere), design system synthesis (.stitch/DESIGN.md), and high-fidelity screen generation/editing via Stitch MCP.
- ▌ UI Hook Wiring Audit · joshuashepherdAudit and fix UI component data wiring — verifies every component renders the correct data by tracing rendered fields back to the type-safe hooks layer. Checks that book thumbnails, author profiles, course cards, video metadata, and all other domain data reach the correct UI locations through the right simplified or custom hooks. Reports status for each component, then fixes only incorrectly wired paths.
- ▌ Asset Generate · joshuashepherdGenerate image assets using Nano Banana 2 (Gemini Flash Image). Use when creating hero images, course covers, book covers, thumbnails, social cards, or any visual asset for the platform.
- ▌ Asset Headshot · joshuashepherdCreate polished headshots and avatars from existing author photos using Nano Banana 2. Handles background replacement, relighting, color grading, cropping, and format adaptation for avatars, team pages, bio cards, OG images, and instructor photos. Use when you need a consistent, on-brand headshot at any size or context.
- ▌ Image Optimize · joshuashepherdOptimize images from the local images repo (convert to WebP, generate responsive variants, resize) and upload to Supabase Storage preserving the multi-tenant folder structure. Use when adding new images, syncing the images repo to storage, or generating responsive sets for existing assets.
- ▌ App Architect · joshuashepherdGenerate a full architectural documentation package for any app, project, or feature — a _PURPOSE.md narrative plus individual page-level specs, navigation, and cross-cutting concerns. Outputs to _docs/ai-studio/ for use in AI Studio or any prototyping tool.
- ▌ Book Pipeline · joshuashepherdOrchestrate the full books pipeline (Phases 01 → 04) for a single book from PDF/EPUB to live on all four vector stores. Routes through `/book-convert` → `/book-validate` → `/book-audit` → optional `/book-fix` → `/book-frontmatter` → `/book-ingest` → `/book-chunk` → `/book-rag-push`, with quality gates between phases. Idempotent — safe to re-run after a fix or copy edit. Use to ingest a brand-new book end-to-end, or to re-process a corrected book through every downstream stage. Trigger phrases: "run the full pipeline", "ingest this book end to end", "phase 01 through 04", "ship this book".
- ▌ Book RAG Push · joshuashepherdPhase 04b of the books pipeline — fan out the canonical `chunks.jsonl` to three vector stores (OpenAI Vector Stores, Gemini File Search, Anthropic Files API) and the in-house pgvector. Idempotent — only re-uploads chunks whose `sha256` changed since the last run; updates `vector_store_ids` map in chapter frontmatter to record which store/version each chapter lives in. Includes the smoke test (same query against all four stores, diff results). Use after `/book-chunk` produces an updated `chunks.jsonl`. Trigger phrases: "push to vector stores", "upload chunks", "phase 04 fan-out", "rag deploy", "smoke test rag".
- ▌ Book Validate · joshuashepherdPhase 01b of the books pipeline — prove no words were lost during PDF/EPUB → Markdown conversion. Runs a five-layer validation harness (word-count diff, fuzzy alignment, 5-gram coverage, embedding similarity, vision LLM judge) comparing the converted MD against the original source. Cheap checks first, semantic checks last, LLM judge only on samples. A chapter that fails any layer goes back to Phase 01 before advancing. Use after `/book-convert`, before `/book-frontmatter`. Trigger phrases: "validate this book", "run the harness", "prove no words were lost", "phase 01b".
- ▌ Course Author · joshuashepherdGenerate course content for a specific section of the transformation loop, following the actual Forgotten Ways course structure, Charter requirements, and Alan Hirsch's voice. Use when writing individual course sections.
- ▌ Course Ingest · joshuashepherdIngest course content from markdown files into the database. Parses manifest + module files, validates, and upserts course_lessons, reflection_questions, discussion_prompts, and exercises.
- ▌ Pathway Audit · joshuashepherdAudit a pathway page against the canonical 12-section architecture. Reports what's missing, what's below spec, what's complete, and what needs improvement for SEO/GEO, voice, and formation quality. Use before publishing a pathway or after adding new sections.
- ▌ Design Section · joshuashepherdDesign a new UI section or component for movemental using DESIGN.md conventions — tonal stacking, primitives, semantic tokens, and the Stitch-informed editorial style.
- ▌ Movemental Ink · joshuashepherdFront-end UI design expert for movemental-ai's "Ink Band" design system — the agent room. Use when building or restyling any screen, primitive, shell zone, or component on the .ink-band-surface to match the existing visual language. Anchors to the .ink-band-surface tokens in src/app/globals.css, the agent-room CSS module (src/components/agent-room/ink-band.module.css), the real component tree (shell/ · screen/ · ink/), and docs/design/INK_BAND_DESIGN_CHAIN.md. Keywords: Ink Band, agent room, warm paper, red margin rule, Playfair display, Inter body, Caveat hand, IBM Plex Mono labels, ink-blue #22409b pen, ink gestures, mast/screen/voice/composer, closed screen set, tonal stacking, hairline borders, staggered reveal, Tailwind v4, shadcn.
- ▌ Grok API · joshuashepherdBuild, refine, and master agents using the xAI Grok API — OpenAI-compatible endpoint with grok-4.20 and grok-4-1-fast model families. Covers function calling, 2M context, multi-agent model, structured output, streaming, and cached inputs. Uses the OpenAI SDK with a base URL override. Fetches live docs from docs.x.ai. Use when building or debugging anything on xAI Grok.
- ▌ Stitch Iterate · joshuashepherdRefine and edit existing Stitch screens using targeted prompts. Uses edit_screens to adjust content, functionality, or states without regenerating from scratch.
- ▌ Studio Testing · joshuashepherd bundleTesting strategy for Supabase Studio. Use when writing tests, deciding what type of test to write, extracting logic from components into testable utility functions, or reviewing test coverage. Covers unit tests, component tests, and E2E test selection criteria.
- ▌ Book Faithful Cleanup · joshuashepherd bundleFaithful PDF/EPUB → movemental chapter Markdown conversion and cleanup: seed-from-corpus, repair-from-db, or convert-from-source; fidelity audit; deterministic repairs; Cursor /loop until READY_FOR_LOAD; Phase 5 citations (book-citations); optional Supabase load. Use when converting a book, running book-faithful-cleanup, pymupdf4llm convert, defect remediation, or iterating cleanup until perfected — never paraphrase source text.
- ▌ Movement Leader Voice · joshuashepherd bundleAuthors the canonical voice and style guide for a movement leader at docs/voice/{SLUG}_VOICE.md — corpus-grounded markers, prose character, failure modes, audit rubric, and sample passages. Use when standing up or refreshing a leader's voice identity, generating docs/voice, running movement-leader-voice, or preparing themes/content in the leader's voice. Distinct from Movemental company voice (movemental-voice) and generic voice-designer.
- ▌ Agent Guardrail · joshuashepherdAdd input/output guardrails to an agent pipeline — validation, content filtering, safety rules. Use when adding safety or quality gates to agents.
- ▌ Setup Streaming · joshuashepherdSet up or debug agent streaming — use when asked to implement SSE streaming, handle ChatKit event types, fix stream stalls or session continuity issues, or wire streaming between an agent API route and a React client.
- ▌ Asset Composite · joshuashepherdCombine multiple generated images into a single composite — grids, collages, mood boards, before/after comparisons, or series overviews. Use when reviewing batches, creating mood boards, or building multi-image marketing assets.
- ▌ Nano Banana Pro · joshuashepherd bundleGenerate or edit images via Gemini 3 Pro Image (Nano Banana Pro). Use when the user asks to create an image, generate a picture, produce AI-generated artwork, edit a photo, compose multiple images, or upscale an image to higher resolution. Supports text-to-image generation, single-image editing, and multi-image composition using the Gemini API.