An eight-stage website-building skill for small businesses working from a phone, with modular design, SEO/GEO, deployment, and domain guidance.
Not a project graveyard—a set of product building blocks
We scanned 59 adjacent project folders and 26 Git repositories, then organized the products, delivery infrastructure, skills, and engineering standards worth carrying forward. Client data stays private; reusable methods, architectures, and product capabilities keep compounding.
What qualifies as reusable
Independently deliverable
A clear user, boundary, and installation or deployment path.
Secure by default
Least privilege, data ownership, HITL, and rollback are design inputs—not later patches.
Customer exit is possible
Customer-owned accounts, exportable data, and no forced maintenance lock-in.
Verifiable
A README, tests, preflight, rendering checks, or human acceptance provides concrete evidence.
Products and engineering assets
These items are regrouped by reuse value. Public repositories are linked; delivery and internal components describe their boundaries without exposing client data, credentials, or private implementation.
Cloudflare delivery foundation
From zero-cost lead-generation sites to bookings, meeting intelligence, private agents, and customer-owned delivery infrastructure.
Browser recording, chunked transcription, structured notes, action items, and calendar sync with passkey access in the user’s Cloudflare account.
A complete booking site for solo service businesses with schedules, passkeys, an owner dashboard, and bilingual UI on the free tier.
Telegram Cloud Agent
A white-label Telegram agent deployed per customer, with multimodal input, durable memory, management check-ins, and strict admin boundaries.
NextAgent Stack
A customer-owned turnkey stack combining sites, bookings, forms, portals, orders, email, support, and Telegram operations.
Crebroker
An edge credential broker and remote MCP gateway that gives agents scoped access to search and model services without exposing upstream keys.
Local-first AI tools
Native macOS products that prioritize privacy, file ownership, and system-level workflows.
A local-first Markdown, EPUB, PDF, and media workspace with wiki links, drawing annotations, and local or remote AI providers.
Hold Fn to speak and release to type into any app, using on-device speech by default or a user-controlled speech endpoint.
OpenClaw Pet
A pixel-art desktop companion that turns agent thinking, replies, and reminders into visible states via local events and Telegram.
Content, knowledge, and education pipelines
Inspectable, repeatable research-to-publish systems instead of one-off prompts.
Loop Engineering
One topic becomes researched copy, images, voice, vertical video, a long poster, and a QC loop, with readable artifacts at every stage.
Obsidian Share Platform
Syncs a local Obsidian course vault into an access-controlled reading platform with assets in R2 and permissions in D1.
Module Library
A governed library of reusable CRM, admin, auth, notification, and website modules with inventories, manifests, versions, and PR review.
Clinic No-Show ML
An end-to-end no-show risk pipeline using synthetic clinic data: features, model comparison, evaluation, calibration, and deployable inference.
CA AI Grant Copilot
A seven-agent framework for Canadian SME AI-transformation grants, covering intake, workflow analysis, program matching, eligibility, one-pagers, intelligence, and red-team review.
AI Film Production Harness
A seven-layer film pipeline for shot lists, reusable character assets, prompt compilation, batch generation, review, assembly, and cost tracking.
Operations and reliable delivery
Observable, reversible, least-privilege operations for servers, agent fleets, and automated outreach.
OpenClaw Fleet
Cross-host OpenClaw and memory-plugin status, compatibility preflight, and daily upgrade reporting through a deliberately read-only skill.
XiaoMai WeChat Automation
Task, template, scheduling, and AI copy workflows for WeChat and WeCom, with desktop execution defaulting to dry-run and human control.
Agent Task Audit Viewer
A local static audit dashboard for NDJSON agent-run logs, showing roles, model visibility, tool use, and token consumption.
Server Operations Toolkit
A unified SSH inventory, health-check, connectivity-diagnostics, and monitoring toolkit for local, cloud, and customer-site services.
Skill library
A skill is not a magic prompt. It is an operating playbook with triggers, workflow, references, verification standards, and safety boundaries.
cf-mobile-site-builder
A complete phone-first Cloudflare site workflow with industry recipes, theme tokens, SEO/GEO, and safe deployment.
typographic-key-visual
Turns a word, name, or news hook into typography-first key-visual prompts and image workflows.
drama-topic-radar
Produces short-drama concepts through market scanning, trope analysis, opportunity detection, and weighted scoring.
personal-ledger
A privacy-first workspace for personal and small-business transactions, receipts, tasks, reminders, and reports.
longevity-pdf-report
A daily longevity, cell-tech, funding, and policy research workflow that produces a polished Chinese PDF.
ssat-lesson-builder
Builds a 10-slide SSAT math lesson and 30-question practice set with programmatic answer and layout checks.
openclaw-fleet
Read-only checks for OpenClaw hosts, plugins, compatibility, and automated upgrade outcomes.
CMS / SEO / Experimentation Standards
Adopted guidance for content modeling, Sanity engineering, SEO/AEO, and experimentation, with upstream attribution preserved.
Inventory scope and disclosure
This page is based on a read-only local inventory completed on September 16, 2026. Maturity labels describe current delivery readiness; they do not imply that every item is open source. Medical and client CRM data, email authorization records, real invoices, student materials, production credentials, server addresses, and private repository paths are excluded. Third-party standards retain their original attribution.
Want to turn one building block into your operating system?
We can start with one narrow, verifiable workflow: prototype it, validate it in real use, and expand only when the evidence supports it.