SOP Pipeline
Every source collected, merged by topic into SOPs Claude can run, and every finished SOP searchable in one place.
SOPs
Finished SOPs open into their full page. Planned ones show what's coming. Status comes from each file's header in the vault.
A bank of 20 scored hooks for one topic, with the top 5 ready to film.
A finished 8 to 10 slide carousel
Captions and short scripts in your voice
One asset turned into every format
A 20 to 40 second text-driven explainer
A product ad from stills and copy
Numbers and charts brought to life
A 30 second avatar reel with captions
A 3 to 8 minute explainer
A consistent character with a reference sheet
Keyword comment delivers a link or freebie
Three questions, scored and logged
Replies for common questions and objections
A live lead campaign with a baseline
A form that qualifies event type and date
Cost per lead and three next actions
Leads Center view a client can read
A scored go or no-go with sources
Who does it, pricing, gaps
Stages, tasks and prompts, like Early Bird
Vetted install of a GitHub repo or plugin
CLAUDE.md, backlog, /next and /wrap
A scroll-driven page, live
Scheduled scraper, D1 data, alerts
A public edition from a master SOP
A page listing every free SOP
A bundle with a checkout
A tool like the Core Needs analyzer
No SOPs match these filters yet.
Data pipeline
From everything saved to the modules that use it. Hover a tool for its role, cost and status.
18 sources
ChatGPT chatsClaude chatsClaude Code project filesInstagram saves and collectionsTikTok favoritesYouTube playlists, Watch Later, likesChrome tabs, bookmarks, reading listGoogle Drive filesGmail newsletters and saved emailsiPhone NotesPhone files: photos, screenshots, voice memosMac files: Downloads, Desktop, DocumentsGitHub stars and reposX bookmarksLinkedIn saved postsTelegram saved messagesWhatsApp notes to selfWebsites, newsletters, misc linksCapture
Connectors pull directly. Everything else arrives as an export.
Extract
Video, audio, PDFs and chat files become text notes.
Triage and merge
Score, cluster by topic, merge each cluster into one SOP, then test it.
Library
SOPs live in the vault as markdown, status in each header.
Run and publish
Proven SOPs become Claude skills and public editions.
Every source
Status is read-only here. To mark a source collected, tick it in "Sources checklist.md" in the vault. Add new sources to _data/sources.yaml.
Settings, Data controls, Export. Arrives by email as a zip with conversations.json.
Settings, Privacy, Export data. JSON of every chat.
CLAUDE.md, decisions.md, backlog.md in each repo. Read directly.
Accounts Center, Download your information, Saved items. Gives links only, not media.
Settings, Download your data. Favorite videos list, if you use TikTok.
Google Takeout, YouTube. Playlists arrive as CSV files of video links.
Open tabs through Claude in Chrome. Bookmarks through Google Takeout or the bookmarks export.
Read directly with the Drive connector.
Read directly with the Gmail connector, filtered by sender or label.
Synced to the Mac through iCloud, read with the Apple Notes connector.
AirDrop or iCloud Drive to the Mac, then into 00_INBOX.
Claude Code or the Mac connector reads the folders in place.
Your stars list, read with the GitHub CLI or the web page.
No official export of bookmarks. Pulled through the browser.
No export. Pulled through the browser.
Telegram Desktop, Export chat history, Saved Messages.
Export chat from the note-to-self thread.
Anything else gets dropped into 00_INBOX as a link note.
Full stack
46 tools from our projects so far. Each module page shows which ones its process uses.
Brain and agents
Knowledge and storage
Capture and scraping
Transcription and voice
Image, avatar and video
Web and UI
Ads, leads and DMs
Hosting, alerts and publishing
Claude Code safety and scale
Other and unconfirmed
What every SOP contains
The template in 07_Templates. Internal parts never appear in public editions.
Title, module, version, status, time, cost, difficulty
Exactly what you have when you finish
Every tool, its job, free or paid
One action per step, with time per stage
Exact text in prompt blocks, [VARIABLES] marked
How to tell a good result from a weak one
A strong and a weak output
Limits, platform rules, common failures
The notes this SOP was merged from
What changed, why, test results
Start here
One module end to end before scaling.
Request the exports
ChatGPT, Claude, Instagram, Google Takeout and TikTok exports take hours to days. Request them all on day one.
Collect every source
Work down the sources checklist in the vault.
Run one module
Extract, triage, cluster, merge and test until one SOP reaches Proven.
Repeat by module
Results from use feed new versions.