armadillo-shepherd

Active router — classifies every request and routes to the correct skill before any response

Model	Source	Category
sonnet	core	Meta

Full Reference

If you think there is even a 1% chance a skill might apply to what you are doing, you ABSOLUTELY MUST invoke the skill.

IF A SKILL APPLIES TO YOUR TASK, YOU DO NOT HAVE A CHOICE. YOU MUST USE IT.

This is not negotiable. This is not optional. You cannot rationalize your way out of this.

Armadillo Shepherd

The orchestration layer. Not documentation — an active router. Every request flows through this routing table before any response.

Mandatory Announcement — FIRST OUTPUT before anything else:

┏━ 🛡 armadillo-shepherd ━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ [one-line description of what request/routing]  ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

No exceptions. Box frame first, then work.

How to Access Skills

In Claude Code: Use the Skill tool. When you invoke a skill, its content is loaded and presented to you — follow it directly. Never use the Read tool on skill files.

State-Aware Routing

Before classifying any request, check for active state:

Read .armadillo/session.json — is there an active session?
Read .armadillo/plans/*.state.json — any in-progress plans?
Classify request weight using signals from the request text

Weight Classification

Weight	Signals	State	Git
Light	Single file, “just”, “quick”, config change	None — no state files	Direct commit
Medium	One feature, 2-5 files, clear scope	Session state only	Single worktree
Heavy	”plan”, “build system”, 5+ tasks, parallelizable	Plan state + session	Multi-branch worktrees

Escalation rule: Start light, escalate if complexity discovered. Never downgrade.

Resume Detection

When session or plan state exists, route by phase:

State Found	Route
`session.phase == "executing"` + plan exists	`executing-plans` (resume)
Plan with `in_progress` tasks, no session	`executing-plans` (rebuild session from plan)
Plan with all tasks `completed`	`finishing-a-development-branch`
No state, request matches “continue” + plan exists	`executing-plans` (resume)

How to Read State

Use .claude/lib/armadillo-state.js and .claude/lib/weight-classifier.js:

readSession(root)       → session object or null
listPlans(root)         → array of plan state objects
readPlanState(root, n)  → specific plan state or null
classifyWeight({request, signals}) → { weight, reason }

Routing decision flow:

readSession() — active session? Route by session.phase
listPlans(root, { status: 'in_progress' }) — active plans? Offer resume
classifyWeight({ request }) — classify new work
Route by weight + phase per tables above

Light Path (fast)

For light tasks, skip all ceremony:

No state files created
No worktree — commit on current branch
Route directly to target skill (TDD, debugging, etc.)
Target: under 2 minutes end-to-end

Routing Table

Classify the request. Invoke the matching skill. No response before invocation.

Creative & Planning

Request	Skill
New feature, idea, anything creative	`brainstorming`
Have a spec/design, need a plan	`writing-plans`
Have a plan, need to execute it	`executing-plans`
New project from scratch, idea to build	`fresh-project`

Implementation

Request	Skill
Implementing ANYTHING	`test-driven-development`
ANYTHING testing (test setup, debugging, choosing tools)	`testing-pimp`
2+ independent tasks	`dispatching-parallel-agents`
Sequential tasks this session	`subagent-driven-development`
Bug, unexpected behavior, mystery	`systematic-debugging`
Test failure diagnosis, flaky tests, test debugging	`systematic-debugging` (quick path)
Post-implementation cleanup, dead code removal	`cleanup`
Dependency audit, update, add packages safely	`deps`
Pre-merge quality gate, merge readiness check	`safe-merge`

Swarms

Swarms are for complex multi-agent coordination — not simple parallel tasks. Use dispatching-parallel-agents for 2-5 independent tasks. Use swarms when tasks need coordination patterns (hierarchy, aggregation, review, or pipelines).

Signal	Use
2-5 independent tasks, no coordination needed	`dispatching-parallel-agents`
6+ tasks needing a coordinator to decompose and delegate	`swarm-hierarchical`
Independent tasks needing aggregated results (hub collects from spokes)	`swarm-star`
Multi-perspective review, consensus needed from multiple agents	`swarm-mesh`
Sequential stages where output of one feeds into the next	`swarm-pipeline`

AI Visibility

Request	Skill
llms.txt, LLM discovery, AI crawlers, llmstxt.org	`llms-txt`
AI info page, structured info for AI agents, /ai-info	`ai-info-page`

Content Creation

Request	Skill
New blog post, article, blog content	`add-blog-post`
Service page, offering page, pricing page	`add-service-page`
Location page, local SEO page, area served	`add-location-page`
Testimonial, review, customer feedback	`add-testimonial`

Client Ops

Request	Skill
Client report, branded report, competitive landscape, service catalog, monthly pulse	`client-reports`
Prospect audit, external site audit, prospect analysis, new client pitch	`prospecting`

Video Pipeline

Request	Skill
Blog-to-video, content-to-video, video pipeline orchestration	`video-pipeline`
Ingest source content, create video brief, read article for video	`read-source-content`
Generate video script, write script, scene-marked script	`generate-video-script`
Storyboard, shot list, visual planning	`build-storyboard`
Catalog footage, scan clips, clip manifest	`ingest-media-clips`
Transcribe clips, Deepgram transcription, word timestamps	`transcribe-clips`
Classify takes, score footage, build edit timeline, EDL	`classify-and-plan-edit`
Execute edit, FFmpeg render, video assembly	`execute-edit`
Final polish, Remotion overlay, thumbnail, multi-format output	`render-and-polish`
YouTube captions, YouTube subtitles, YouTube transcript, yt-dlp	`yt-dlp`
Download YouTube video, extract YouTube audio	`yt-dlp`

Ad Assets

Request	Skill
Classify ad images, Gemini Vision assessment, ad eligibility	`classify-ad-assets`
Face-center crop, smart crop, subject centering	`face-center-crop`
Deduplicate images, remove duplicate assets	`deduplicate-assets`
Generate ad variants, platform sizes, resize for ads	`generate-ad-variants`
Sync to Cloudinary, upload ad variants	`sync-to-cloudinary`
Full ad asset pipeline, classify to sync	`ad-asset-pipeline`

Completion & Delivery

Request	Skill
Claiming work is done	`verification-before-completion`
Branch complete, need to ship	`finishing-a-development-branch`
Creating/writing a PR	`writing-prs`
Requesting code review	`requesting-code-review`
Received review feedback	`receiving-code-review`

Git & Workspace

Request	Skill
Feature work needs isolation	`worktree`
Set up git workflow, branch protection, version bump	`git-setup`

Armadillo Meta

Request	Skill
Install armadillo, onboard, init, setup, sync	`armadillo-sync`
Update armadillo, upgrade, check for updates	`armadillo-sync`
Plugin health check, doctor, diagnose plugins	`armadillo-doctor`
Create any armadillo artifact (skill, pack, agent, rule, hook)	`armadillo-maker`
Update existing skill, refresh reference docs, update agent	`armadillo-maker`
Create a new skill (lightweight, no full pipeline)	`writing-skills`
Document an external API (lightweight, no full pipeline)	`writing-reference-skills`
MCP servers, MCP configuration, .mcp.json, MCP troubleshooting	`mcp-management`
Claude Code hooks, hook events, hookSpecificOutput, hook debugging	`claude-code-hooks`
Claude Code agents, subagent frontmatter, agent memory, agent isolation	`claude-code-agents`

Greenfield Project

Request	Skill
”I want to build…”, new project from idea	`fresh-project`
Scaffold from stack.json	`scaffold`
Stack/technology recommendation	Load `stack-recommender` reference

Testing Tools

Request	Skill
E2E / browser testing (Playwright)	`playwright`
Browser automation / Chrome	`puppeteer`
Component or E2E (Cypress)	`cypress`
Unit / component testing	`vitest`

Frontend

Request	Skill
ANYTHING frontend/UI/styling/design/component testing (style, layout, CSS, theme, visual, component, UI modernize)	`frontend-pimp`
Styling / CSS, Tailwind config, utility classes	`frontend-pimp` → routes to `tailwind-css`
UI components, shadcn/ui, Radix primitives	`frontend-pimp` → routes to `shadcn-ui`
Next.js / App Router, RSC, API routes	`frontend-pimp` → routes to `nextjs`
Astro / content sites, islands	`frontend-pimp` → routes to `astro`
React SPA, Vite setup, client-side app	`frontend-pimp` → routes to `react-vite`
SvelteKit, Svelte routing	`frontend-pimp` → routes to `sveltekit`
Animations (React), mount/unmount, gestures	`frontend-pimp` → routes to `framer-motion`
Scroll / timeline animations, GSAP plugins	`frontend-pimp` → routes to `gsap`
Responsive / mobile-first, viewport, container queries	`frontend-pimp` → routes to `responsive-design`
Accessibility / WCAG, ARIA, keyboard nav	`frontend-pimp` → routes to `accessibility`
Design tokens, CSS architecture, theming, CSS debugging	`frontend-pimp` → routes to `design-system`
UI vibes, style guide, modernize UI, aesthetic direction	`frontend-pimp` → routes to `ui-craft`
Storybook, component stories, CSF, visual dev environment	`frontend-pimp` → routes to `storybook`
Visual regression, screenshot testing, Chromatic, Argos	`frontend-pimp` → routes to `visual-regression`
Component testing, @testing-library, DOM queries, render tests	`frontend-pimp` → routes to `testing-library`

UX/UI (AI-Assisted Design-to-Code)

Request	Skill
ANYTHING AI-assisted UI generation (21st.dev, Stitch, Figma MCP, component registry)	`ux-ui-pimp`
21st.dev Magic MCP, search UI components, install component from registry	`ux-ui-pimp` → routes to `magic-21st`
Stitch MCP, generate full-page design, AI page layout from brief	`ux-ui-pimp` → routes to `stitch-design`
Figma MCP, read Figma file, extract code from Figma, Figma Dev Mode	`ux-ui-pimp` → routes to `figma-sync`
Component library, component-library.json, manage component registry	`ux-ui-pimp` → routes to `component-library`
End-to-end UI generation pipeline, brief to code, design-to-code workflow	`ux-ui-pimp` → routes to `generate-workflow`
”find me a button component”, “get a navbar from 21st”, “design a landing page”	`ux-ui-pimp`
Compare UI generation tools, “which MCP tool for UI?”	`ux-ui-expert` agent

Backend & API

Request	Skill
ANYTHING backend API (Hono, Express, tRPC, REST design)	`backend-pimp`
Build API with Hono, Cloudflare Workers API	`backend-pimp` → routes to `hono`
Build API with Express, Express middleware, Express routing	`backend-pimp` → routes to `express`
Build type-safe API with tRPC, tRPC setup	`backend-pimp` → routes to `trpc`
Design REST API, REST resource naming, HTTP methods guide	`backend-pimp` → routes to `rest-api-patterns`

Database & ORM

Request	Skill
ANYTHING database or ORM	`database-pimp`
Supabase database, Supabase realtime, Supabase RLS	`supabase`
MongoDB queries, Mongoose, MongoDB aggregation	`mongodb`
Redis caching, Upstash rate limiting, background queues	`redis-upstash`
Drizzle ORM schema, Drizzle queries, Drizzle migrations	`drizzle`
Prisma schema, Prisma client queries, Prisma migrations	`prisma`
PostgreSQL vectors, pgvector, similarity search, embeddings	`postgresql-pgvector`
Turso, libSQL, embedded replicas, edge database, Turso CLI, Turso vector search, Turso branching, Turso pricing, Turso AI agents, per-agent database	`turso`

Authentication

Request	Skill
ANYTHING authentication (Auth.js, Clerk, Supabase Auth, Better Auth)	`auth-pimp`
Set up Auth.js, NextAuth, OAuth with Next.js	`auth-pimp` → routes to `authjs`
Set up Clerk auth, Clerk organizations	`auth-pimp` → routes to `clerk`
Supabase auth, magic links, Supabase RLS with auth	`auth-pimp` → routes to `supabase-auth`
Better Auth, better-auth sessions, TypeScript auth, Drizzle auth adapter	`auth-pimp` → routes to `better-auth`

Infrastructure & Deployment

Request	Skill
ANYTHING deployment or infrastructure (Vercel, Cloudflare, Docker, CI/CD, Railway)	`deploy-pimp`
Deploy to Vercel, Vercel environment variables, Vercel functions	`deploy-pimp` → routes to `vercel`
Deploy to Cloudflare, Cloudflare Workers, Cloudflare D1	`deploy-pimp` → routes to `cloudflare-pages-workers`
Dockerize app, Docker Compose, container deployment	`deploy-pimp` → routes to `docker`
Set up CI/CD, GitHub Actions workflow, automated testing	`deploy-pimp` → routes to `github-actions`
Railway deploy, Railway CLI, Railway Postgres	`deploy-pimp` → routes to `railway`
GitHub API, GitHub Issues, GitHub Releases, GitHub Pages	`deploy-pimp` → routes to `github-platform`

Developer Experience

Request	Skill
Zod validation schema, parse API input, validate form data	`zod`
React form, form validation, multi-step form	`react-hook-form`
Client state management, Zustand store, global state	`zustand`
Data fetching, server state, useQuery, useMutation	`tanstack-query`
SWR data fetching, React stale-while-revalidate	`swr`
Error tracking, Sentry setup, error monitoring	`sentry`
Analytics, feature flags, PostHog setup	`posthog`
ESLint setup, Prettier config, code formatting	`eslint-prettier`
Monorepo setup, Turborepo pipeline, workspace packages	`turborepo`

APIs & Services

Request	Skill
Google Analytics 4	`ga4-api`
Google Ads	`google-ads-api`
Google Search Console	`google-search-console-api`
Google Business Profile	`google-business-profile-api`
Google Places	`google-places-api`
Lighthouse / PageSpeed	`lighthouse-api`
YouTube	`youtube-data-api`
ANYTHING payments (Stripe, Square, POS, invoices)	`payments-pimp`
Stripe / payments	`payments-pimp` → routes to `stripe-api`
Square payments, orders, checkout, invoices, subscriptions	`payments-pimp` → routes to `square-payments`
Square catalog, items, inventory, stock	`payments-pimp` → routes to `square-catalog`
Square Terminal, in-person payments, device pairing	`payments-pimp` → routes to `square-terminal`
Square loyalty, gift cards, bookings, appointments	`payments-pimp` → routes to `square-engagement`
Square API auth, OAuth, SDK setup, webhooks, error codes	`payments-pimp` → routes to `square-api-reference`
Neon / Postgres	`neon`
Neon backup, database backup, pg_dump, pre-migration backup	`neon`
Acuity / Squarespace Scheduling, appointments, availability, booking flows, webhook signatures	`scheduling`
Google Tag Manager, GTM, data layer, server-side tagging	`google-tag-manager`

Content & Creative

Request	Skill
ANYTHING file uploads or storage	`storage-pimp`
ANYTHING CMS (Sanity, Payload, Keystatic, headless CMS)	`cms-pimp`
Sanity CMS, GROQ queries, Sanity Studio, headless CMS	`cms-pimp` → routes to `sanity`
Payload CMS, code-first CMS, Payload collections	`cms-pimp` → routes to `payload`
Keystatic CMS, git-based CMS, Keystatic + Astro	`cms-pimp` → routes to `keystatic`
MDX, remark plugins, rehype plugins, MDX components	`mdx`
Send emails, transactional email, Resend API	`resend`
Email templates, React email components	`react-email`
File uploads, image upload, Uploadthing	`uploadthing`
Object storage, S3 upload, R2 storage, presigned URLs	`s3-cloudflare-r2`
ASCII art / CLI visuals	`ascii-art`
Audio transcription	`deepgram-transcription`
Programmatic video	`remotion`
Render Remotion video, encoding, output formats	`render-video`
Supercut, highlight reel, clip compilation	`create-supercut`
Remotion composition template, parameterized video	`create-template`
Ingest content, media pipeline, transcription prep	`ingest-content`
FFmpeg, video transcoding, audio processing, HLS packaging	`ffmpeg`
Cloudinary uploads, image/video optimization, CDN	`cloudinary`
Website migration (Duda→Astro)	`duda-to-astro-migration`

Web Scraping & Data

Request	Skill
Web scraping, crawl website, extract web data, Firecrawl	`firecrawl`
Convert website to markdown, LLM-ready data, structured extraction	`firecrawl`
Firecrawl MCP server, Firecrawl CLI, Browser Sandbox	`firecrawl`

Brand

Request	Skill
ANYTHING brand-related (brand audit, build, voice, assets, guidelines, export, compliance)	`brand-pimp`
Brand audit, what brand assets exist, brand gap report	`brand-pimp` → routes to `brand-discovery`
Brand interview, build brand knowledge, process brand docs, transcribe brand audio	`brand-pimp` → routes to `brand-knowledge-builder`
Organize brand assets, set up brand.json, organize logos/images	`brand-pimp` → routes to `brand-asset-organizer`
Export brand package, generate brand PDF, create asset zip, client deliverables	`brand-pimp` → routes to `brand-export`
Check brand compliance, is this on-brand, brand content review	`brand-pimp` → routes to `brand-compliance`
Audio transcription for brand interviews	`deepgram-transcription`

Merch

Request	Skill
ANYTHING merch-related (merch, merchandise, t-shirt, sticker, apparel, blank, mockup, gear)	`merch-pimp`
Printful, POD, sell merch, online store, product listing	`merch-pimp` → routes to `printful-pod`
Printful API, Printful auth, Printful endpoints, Printful webhook	`merch-pimp` → routes to `printful-api`
Die-cut sticker, sticker design, sticker file, kiss-cut, sticker forge	`merch-pimp` → routes to `sticker-forge`
Premium blanks, team gear, expo merch, DTG, DTF, screen print	`merch-pimp` → routes to `premium-gear`
Generate mockup, product mockup, t-shirt mockup, sticker mockup	`merch-pimp` → routes to `mockup-generator`
Merch quality check, visual QA, inspect output	`merch-pimp` → routes to `merch-qa`

macOS Automation

Request	Skill
ANYTHING macOS automation (automate macOS, desktop control, GUI scripting, AppleScript, Peekaboo, osascript)	`macos-pimp`
Peekaboo, screenshot, screen capture, see/click/type, window/menu/app control, MCP server	`macos-pimp` → routes to `peekaboo`
AppleScript execution, JXA, scripting tips, script knowledge base, run script by ID	`macos-pimp` → routes to `macos-automator`
AppleScript syntax, JXA language, app dictionaries, applescript-mcp, SSH remote	`macos-pimp` → routes to `applescript-reference`
Multi-step macOS workflow, automation chain, .peekaboo.json, See→Act→Verify	`macos-pimp` → routes to `macos-workflow`
Generate AppleScript, write JXA, validate script, forge script	`macos-pimp` → routes to `applescript-forge`
Build macOS AI agent, NL desktop automation, agent config, MCP tool chain	`macos-pimp` → routes to `macos-agent-builder`
Compare macOS automation tools, “which tool for X”, permission debugging	`macos-automation-expert` agent

Google CLI

Request	Skill
ANYTHING Google CLI (gogcli, goplaces, gcloud, Google Ads CLI, gog, MCC)	`google-pimp`
Gmail, Calendar, Drive, Docs, Sheets, Workspace via CLI, gog command	`google-pimp` → routes to `gogcli`
Places search, nearby, directions, geocoding CLI, goplaces command	`google-pimp` → routes to `goplaces`
Google Ads GAQL, MCC, campaign CLI, gaql, mcc-gaql, google-ads-mcp	`google-pimp` → routes to `google-ads-cli`
gcloud, Compute, GKE, Cloud Run, IAM, GCP infra	`google-pimp` → routes to `gcloud`
Workspace automation pipeline (email → sheet → calendar)	`google-pimp` → routes to `google-workspace-workflow`
MCC reporting, cross-account ads health check	`google-pimp` → routes to `google-ads-workflow`
Set up google-cli.json, add client context, switch GOOGLE_CLI_CONTEXT, validate auth	`google-pimp` → routes to `google-accounts`
Compare Google CLI tools, “which Google tool for X”	`google-cli-expert` agent

Request	Skill
ANYTHING Telegram (bots, Mini Apps, Stars, inline, channels, webhooks, login)	`telegram-pimp`
Telegram Bot API, sendMessage, bot methods, update handling, rate limits	`telegram-pimp` → routes to `telegram-bot-api`
Telegram Mini Apps, WebApp SDK, initData validation, CloudStorage	`telegram-pimp` → routes to `telegram-mini-apps`
Telegram Stars, XTR payments, in-app purchases, subscriptions	`telegram-pimp` → routes to `telegram-stars`
Telegram inline bots, inline queries, keyboards, callback queries	`telegram-pimp` → routes to `telegram-inline`
Telegram channels, groups, admin rights, member management	`telegram-pimp` → routes to `telegram-channels`
Telegram webhooks, setWebhook, getUpdates, BotFather, secret_token	`telegram-pimp` → routes to `telegram-webhooks`
Telegram Login Widget, auth hash validation, user data	`telegram-pimp` → routes to `telegram-login`

Paid Ads

Request	Skill
ANYTHING paid ads (Meta Ads, Pinterest Ads, CAPI, ad campaigns, ad creative)	`ads-pimp`
Meta Ads campaigns, ad sets, ad creative, Meta Marketing API	`ads-pimp` → routes to `meta-ads`
Meta custom audiences, lookalike audiences, audience management	`ads-pimp` → routes to `meta-audiences`
Meta Conversions API, CAPI, server-side events, event dedup	`ads-pimp` → routes to `meta-conversions`
Meta Graph API reference, API version, rate limits, auth flows	`ads-pimp` → routes to `meta-api-reference`
Pinterest Ads, Pinterest campaigns, promoted pins	`ads-pimp` → routes to `pinterest-ads`
Pinterest API reference, OAuth, token refresh, rate limits	`ads-pimp` → routes to `pinterest-api-reference`
Verify Meta auth tokens, debug Meta API permissions	`ads-pimp` → routes to `verify-meta-auth`

Request	Skill
ANYTHING organic social or local business listings	`social-pimp`
Facebook Page posts, organic Facebook publishing, page management	`social-pimp` → routes to `meta-pages`
Facebook Page insights, page reviews, page info	`social-pimp` → routes to `meta-pages`
Instagram organic posts, Reels, Stories, carousels	`social-pimp` → routes to `instagram-api`
Instagram Graph API, IG content publishing, Instagram business account	`social-pimp` → routes to `instagram-api`
Google Business Profile, GBP listing, manage Google listing	`social-pimp` → routes to `google-business-profile-api`
Google Maps listing, Google reviews, business hours on Google	`social-pimp` → routes to `google-business-profile-api`
Post to Facebook and Instagram	`social-pimp` (chains both)
Manage reviews across platforms	`social-pimp` (chains GBP + meta-pages)

SEO

Request	Skill
ANYTHING SEO	`seo-pimp`
Full SEO audit, technical SEO, site optimization	`seo-audit`
Quick SEO pulse check, page-level SEO analysis	`seo-pulse`
Local SEO audit, Google Business Profile optimization, NAP consistency	`local-seo-audit`
Backlink analysis, link profile audit, toxic links	`link-analysis`
Search rank tracking, SERP monitoring, keyword positions	`search-rank`
Full site report, SEO dashboard, performance summary	`site-report`
Schema, structured data, JSON-LD, rich results	`schema-markup`
Reviews, reputation, review generation, response management	`review-management`
Content strategy, topic clusters, content calendar	`content-strategy`
AI visibility, LLM citations, AI search, llms.txt	`ai-visibility`

CRO

Request	Skill
ANYTHING CRO (conversion optimization, A/B testing, heatmaps, tracking)	`cro-pimp`
CRO audit, conversion optimization, conversion rate	`cro-pimp` → routes to `cro-audit`
A/B test, experiment, split test, multivariate test	`cro-pimp` → routes to `ab-testing`
Landing page optimization, hero, CTA, form optimization	`cro-pimp` → routes to `landing-page-cro`
Heatmaps, session recordings, rage clicks, Microsoft Clarity	`cro-pimp` → routes to `microsoft-clarity`
Server-side tracking, CAPI fan-out, event dedup, /api/track	`cro-pimp` → routes to `server-side-tracking`

Performance

Request	Skill
Core Web Vitals, page speed, LCP/INP/CLS fix	`web-performance`
CrUX field data, Chrome UX Report, real-user metrics	`crux-api`

Python

Request	Skill
ANYTHING Python web dev	`python-pimp`
Django project, Django views, Django URL routing	`django`
Django ORM, Django models, QuerySet, migrations	`django-orm`
Django authentication, Django permissions, user management	`django-auth`
FastAPI endpoints, Pydantic models, async Python API	`fastapi`

AI & Frontier

Request	Skill
ANYTHING AI models or inference	`ai-pimp`
AI chat interface, LLM streaming, AI SDK tool calling	`vercel-ai-sdk`
Claude API, Anthropic messages, tool use	`anthropic-api`
Google Gemini, @google/genai SDK, Gemini models	`google-genai`
OpenAI API, GPT models, Chat Completions, Responses API	`openai-api`
fal.ai, FLUX, AI image/video generation, GPU inference	`fal-ai`
Mobile app, React Native, Expo setup	`expo-react-native`

Data Quality

Request	Skill
Hardcoded business info, centralize NAP data, business.json, contact info scattered	`nap-ninja`
Hardcoded API keys, organize .env, environment variables, secrets in source	`env-ninja`

Workflows

Request	Skill
New client audit, full site audit, onboard client site, quarterly deep-dive	`master-performance-assessment` (standard mode)
Monthly retainer check, monthly report, retainer deliverable, quick pulse	`master-performance-assessment` (quick mode)
Full cross-domain audit, competitor benchmarking, deep performance assessment, all APIs	`master-performance-assessment` (deep mode)
Performance history, trend deltas, snapshot management, performance-history.json	`performance-memory`
CRO sprint, conversion experiment cycle, CRO iteration	`cro-sprint`
Local business growth, local SEO pipeline, service area growth	`local-growth`
Set up tracking, analytics stack, CAPI setup, tracking foundation	`tracking-foundation`

OpenCode

Request	Skill
ANYTHING OpenCode-related (config, agents, tools, MCP, plugins, CLI, SDK, ACP, GitHub)	`opencode-pimp`
OpenCode configuration, opencode.json, providers, models, themes, keybinds, formatters, LSP	`opencode-pimp` → routes to `opencode-config`
OpenCode agents, Build/Plan mode, custom agents, subagents	`opencode-pimp` → routes to `opencode-agents`
OpenCode tools, permissions, custom tools, allow/deny/ask	`opencode-pimp` → routes to `opencode-tools`
OpenCode MCP servers, plugins, hooks, commands, skills, rules, AGENTS.md	`opencode-pimp` → routes to `opencode-extensions`
OpenCode CLI, TUI, server mode, SDK, ACP, GitHub integration, Zen	`opencode-pimp` → routes to `opencode-cli`

Research & Recon

Request	Skill
ANYTHING deep research/recon (competitor, audience, content, person, web research)	`deep-recon-pimp`
Web research, fact-finding, multi-source verification	`deep-recon-pimp` → routes to `web-research`
Competitor analysis, market research, competitive landscape	`deep-recon-pimp` → routes to `competitor-analysis`
Audience research, demographics, buyer personas, JTBD	`deep-recon-pimp` → routes to `audience-analysis`
Content research, topic authority, keyword clusters	`deep-recon-pimp` → routes to `content-research`
Person research, public figure, bio, social presence	`deep-recon-pimp` → routes to `person-research`
Compare research approaches, “how should I research X”	`recon-expert` agent

Desktop

Request	Skill
Tauri desktop app, Tauri mobile, Tauri IPC, Tauri plugins	`tauri`
Rust for Tauri, Rust basics, cargo basics, ownership/borrowing	`rust-for-tauri`
Compare desktop frameworks, “Tauri vs Electron”	`tauri-expert` agent

Context Engineering

Request	Skill
Context degradation, context window management, token budgeting	`context-health`
LLM-as-judge, evaluation rubrics, scoring agent outputs	`llm-evaluation`
Tool output management, observation masking, context offloading	`observation-masking`

Scheduling

Request	Skill
ANYTHING scheduling (schedule task, cron, recurring job, “run X at Y time”, list schedules, schedule status)	`armadillo-scheduler`

Content Pumper

Request	Skill
ANYTHING content creation, content pipeline, content automation	`content-pumper-pimp`
Write blog post, create article, generate content	`content-pumper-pimp` → routes to `content-writer`
Research topic for content, content research brief	`content-pumper-pimp` → routes to `content-researcher`
Generate blog illustrations, content visuals, hero image	`content-pumper-pimp` → routes to `content-illustrator`
Content quality check, assess content, content QA	`content-pumper-pimp` → routes to `content-qa`
Trending topics, what’s hot, social buzz, trend scan	`topic-brain-pimp` → routes to `trend-scanner`
Topic sentiment, polarization map, sides of debate	`topic-brain-pimp` → routes to `sentiment-mapper`
Score topics, topic ranking, topic leaderboard	`topic-brain-pimp` → routes to `topic-scorer`
Topic memory, topic history, content-topics.json	`topic-brain-pimp` → routes to `topic-memory`
Schedule content, content cron, autonomous content	`armadillo-scheduler` (with content-pumper presets)
Show topics, browse topics, content-pumper topics	`topic-brain-pimp` (manual mode)

Domain & DNS

Request	Skill
ANYTHING DNS/domain-related (DNS records, domain registration, transfers, WHOIS, nameservers, email auth)	`dns-pimp`
DNS ops without specifying provider, “add A record for client.com”	`dns-pimp` → routes to `dns-manager`
Cloudflare DNS, CF proxy, Cloudflare zone, wrangler DNS	`dns-pimp` → routes to `cloudflare-dns`
GoDaddy DNS, GoDaddy domain, GoDaddy records	`dns-pimp` → routes to `godaddy-dns`
Namecheap DNS, Namecheap domain, Namecheap records	`dns-pimp` → routes to `namecheap-dns`
Name.com DNS, Name.com domain, Name.com records	`dns-pimp` → routes to `namecom-dns`
Squarespace DNS, Squarespace domain, Google Domains migration	`dns-pimp` → routes to `squarespace-dns`
Set up domains.json, configure DNS provider mapping	`dns-pimp` → routes to `dns-manager`
DNS propagation check, nameserver check, dig query	`dns-pimp` → routes to `dns-manager`
Email auth (SPF, DKIM, DMARC) setup or verification	`dns-pimp` → routes to `dns-manager`
Compare DNS providers, “which registrar should I use”	`dns-expert` agent

Mail

Request	Skill
ANYTHING physical mail (certified mail, virtual mailbox, address verification)	`mail-pimp`
Lob API, send certified mail, address verification, letter templates	`mail-pimp` → routes to `lob`
PostScanMail, virtual mailbox, mail scanning, forwarding rules	`mail-pimp` → routes to `postscanmail`
Compare mail services, “Lob vs direct USPS”	`mail-expert` agent

Banking

Request	Skill
ANYTHING banking data (bank accounts, transactions, balances, identity)	`teller`
Teller API, bank account linking, mTLS auth, transaction history	`teller`
ACH payments, payment initiation via Teller	`teller`
Compare banking APIs, “Teller vs Plaid vs MX”	`teller-expert` agent

Workflows

First-class skill chains for common multi-step tasks. When a request matches a workflow, guide through the full chain — don’t stop at the first skill.

Feature Workflow

brainstorming → [domain pimp?] → writing-plans → executing-plans → verification → finishing-branch → writing-prs
     WHAT          CHECK              HOW             DO IT           PROVE IT        SHIP IT         TELL IT

After brainstorming approves a design, always check for a domain pimp before invoking writing-plans. If the design maps to a domain with a pimp orchestrator, invoke the pimp — it handles routing to the right sub-skill AND chains forward. Only invoke writing-plans directly when no pimp applies.

Domain	Invoke instead of writing-plans
Frontend, UI, styling, components	`frontend-pimp`
AI-assisted UI generation, 21st.dev, Stitch, Figma MCP	`ux-ui-pimp`
Brand assets, voice, guidelines	`brand-pimp`
Merch, POD, stickers, mockups	`merch-pimp`
macOS automation, AppleScript, Peekaboo	`macos-pimp`
DNS, domains, nameservers, email auth	`dns-pimp`
Google CLI (gogcli, gcloud, Ads CLI)	`google-pimp`
OpenCode config, agents, extensions	`opencode-pimp`
Paid ads (Meta Ads, Pinterest Ads, CAPI)	`ads-pimp`
Organic social (Facebook Pages, Instagram, Google listing)	`social-pimp`
Deployment (Vercel, Cloudflare, Docker, Railway)	`deploy-pimp`
Payments (Stripe, Square, POS)	`payments-pimp`
Authentication (Auth.js, Clerk, Supabase Auth, Better Auth)	`auth-pimp`
Physical mail, certified mail, virtual mailbox	`mail-pimp`
CMS (Sanity, Payload, Keystatic)	`cms-pimp`
CRO (audits, A/B tests, heatmaps, tracking)	`cro-pimp`
Backend API (Hono, Express, tRPC, REST)	`backend-pimp`
Anything else	`writing-plans` ← default

Phase	Skill	TDD Role
Design	`brainstorming`	None — pure design thinking
Domain check	pimp (if applicable)	Routes to domain sub-skill
Plan	`writing-plans`	RED/GREEN/REFACTOR baked into every task step
Execute	`executing-plans`	Invokes `test-driven-development` per task
Verify	`verification-before-completion`	Runs full suite, confirms coverage
Ship	`finishing-a-development-branch`	Final quality gate
PR	`writing-prs`	Documents test plan in PR body

Bug Workflow

systematic-debugging → test-driven-development → verification → finishing-branch
      UNDERSTAND              FIX                    PROVE             SHIP

Refactor Workflow

brainstorming → writing-plans → subagent-driven-development → verification → finishing-branch
    WHAT             HOW              DO IT (parallel)           PROVE           SHIP

Maker Workflow

armadillo-maker → writing-plans → subagent-driven-development → verification → finishing-branch → writing-prs
   CLASSIFY         PLAN              BUILD + REGISTER           PROVE           SHIP              TELL
   RECON
   RESEARCH
   DESIGN

Phase	Skill	Role
Classify + Recon + Research + Design	`armadillo-maker`	Owns first 4 phases internally
Plan	`writing-plans`	TDD implementation plan from design doc
Build	`subagent-driven-development`	TDD execution, delegates to writing-skills / writing-reference-skills
Register	`armadillo-maker` (resumes)	armadillo.json, sync-all, shepherd, hooks.json
Prove	`verification-before-completion`	Full suite + behavioral testing
Ship	`finishing-a-development-branch`	PR → merge → cleanup

Client Audit Workflow

seo-audit / cro-audit / local-seo-audit → client-reports
            ANALYZE                           DELIVER

Research Workflow

deep-recon-pimp → brainstorming → writing-plans → executing-plans
    RESEARCH          DESIGN          PLAN              BUILD

Desktop App Workflow

brainstorming → tauri + rust-for-tauri → writing-plans → executing-plans
    DESIGN          DOMAIN SKILLS          PLAN              BUILD

Prospect Intelligence Workflow

deep-recon-pimp → prospecting → client-reports
    RESEARCH         AUDIT          DELIVER

Use when preparing richer prospect intelligence before a client audit. Deep research feeds into the prospect crawl for more informed analysis.

Content-to-Video Workflow

content-pumper-pimp → video-pipeline → render-and-polish
    WRITE               PRODUCE         DELIVER

Use when converting published blog content into video format. Content-pumper creates the article, video-pipeline transforms it into a video script and edit.

Research-to-Content Workflow

deep-recon-pimp → content-pumper-pimp → [delivery skill]
    RESEARCH          WRITE + QA          PUBLISH

Use when research feeds directly into content creation. Deep recon produces the brief, content-pumper consumes it for writing. The content-researcher step inside content-pumper reuses the existing research brief instead of re-researching.

Hard Rules

Never respond before invoking the target skill
No summarizing, planning to invoke, or explaining what you’re about to do
If two skills apply, invoke the FIRST one — it chains to the next
If unclear, ask ONE clarifying question, then route
Skill not in the table? Respond directly — no skill needed
If the project directory is empty/near-empty (no package.json, no src/, no framework config) AND the user describes something to build → route to fresh-project

Red Flags

These thoughts mean STOP — you’re rationalizing:

Thought	Reality
”This is just a simple question”	Questions are tasks. Check the routing table.
”I need more context first”	Skill check comes BEFORE clarifying questions.
”Let me explore the codebase first”	Skills tell you HOW to explore. Route first.
”I can check git/files quickly”	Files lack conversation context. Route first.
”Let me gather information first”	Skills tell you HOW to gather information.
”This doesn’t need a formal skill”	If a skill exists, use it.
”I remember this skill”	Skills evolve. Read current version.
”This doesn’t count as a task”	Action = task. Check the routing table.
”The skill is overkill”	Simple things become complex. Use it.
”I’ll just do this one thing first”	Route BEFORE doing anything.
”This feels productive”	Undisciplined action wastes time. Skills prevent this.
”I know what that means”	Knowing the concept ≠ using the skill. Invoke it.

Skill Priority

When multiple skills could apply, use this order:

Process skills first (brainstorming, debugging) — these determine HOW to approach the task
Implementation skills second (TDD, frontend tools) — these guide execution

“Let’s build X” → brainstorming first, then implementation skills. “Fix this bug” → debugging first, then domain-specific skills.

Skill Types

Rigid (TDD, debugging): Follow exactly. Don’t adapt away discipline.

Flexible (patterns, frontend): Adapt principles to context.

The skill itself tells you which.

User Instructions

Instructions say WHAT, not HOW. “Add X” or “Fix Y” doesn’t mean skip workflows.

armadillo-shepherd

Armadillo Shepherd

How to Access Skills

State-Aware Routing

Weight Classification

Resume Detection

How to Read State

Light Path (fast)

Routing Table

Creative & Planning

Implementation

Swarms

AI Visibility

Content Creation

Client Ops

Video Pipeline

Ad Assets

Completion & Delivery

Git & Workspace

Armadillo Meta

Greenfield Project

Testing Tools

Frontend

UX/UI (AI-Assisted Design-to-Code)

Backend & API

Database & ORM

Authentication

Infrastructure & Deployment

Developer Experience

APIs & Services

Content & Creative

Web Scraping & Data

Brand

Merch

macOS Automation

Google CLI

Telegram

Paid Ads

Organic Social & Listings

SEO

CRO

Performance

Python

AI & Frontier

Data Quality

Workflows

OpenCode

Research & Recon

Desktop

Context Engineering

Scheduling

Content Pumper

Domain & DNS

Mail

Banking

Workflows

Feature Workflow

Bug Workflow

Refactor Workflow

Maker Workflow

Client Audit Workflow

Research Workflow

Desktop App Workflow

Prospect Intelligence Workflow

Content-to-Video Workflow

Research-to-Content Workflow

Hard Rules

Red Flags

Skill Priority

Skill Types

User Instructions