Cowork Setup – July 11, 2026

The Fertile Soil Operations: Getting Claude Running – July 11, 2026

Overview

The Fertile Soil freely shares our knowledge. With that in mind, we are sharing our workflow for today.


What We Did

Task Outcome
Cowork role setup Picked wrong role first (Product Management), corrected once real context came up
Plugin install Installed small-business + operations – no dedicated nonprofit plugin exists yet
Compliance research Verified current CA/IRS filing deadlines for a young 501(c)(3)
Launch roadmap Produced a 3-phase roadmap doc (Now / Next / Later)
Website audit Logged into wp-admin, sorted real content from template junk across 11 pages, 2 posts
Site cleanup Real copy for Contact, Volunteer, and FAQ; trashed duplicate Donor Dashboard page; restored a misjudged post
Official info Added EIN and 501(c)(3) status to the site footer and About page
Navigation Added a Stories link so the blog is reachable from the main menu
Real photos Uploaded 30 real photos; swapped stock/broken images on the Contact, FAQ, Stories, and Volunteer pages
Library page Taken offline (draft) since there is no content yet; removed from navigation
Menu cleanup Removed a dead Events link; fixed the footer menu, which had a broken link and a mislabeled link pointing to the wrong page

Part 1: Getting Cowork Set Up

Went through Cowork’s onboarding flow, which starts by asking your role so it can suggest matching plugins. First pass picked "Product management" – wrong context entirely. Once I explained The Fertile Soil is a nonprofit, Claude re-searched the plugin marketplace and came back with small-business (cash flow, contract review, hiring, onboarding walkthrough) and operations (SOPs, RACI, risk assessment) instead. Neither is nonprofit-specific – that plugin doesn’t exist yet in the marketplace – but small-business has a built-in "getting started" flow that’s close enough.

Lesson: the role picker is a starting point, not a commitment. Worth correcting immediately if the first suggestion doesn’t fit, rather than working around a mismatched plugin.


Part 2: From "Product Manager" to Nonprofit Founder

Ran the small-business:smb-onboard skill, which is meant to connect tools like QuickBooks and HubSpot – none of which matter yet since we’re pre-revenue and pre-space. Skipped the connector push entirely and went straight to what’s actually useful right now: a roadmap.

Facts on the table going in: California 501(c)(3), EIN issued, bank account open, mission is a 10,000 sq ft free community makerspace in Oakland (wood shop, metal shop, music, computer lab), website live but not finished, physical space not yet leased.


Part 3: The Launch Roadmap

Before writing anything, had Claude verify current CA compliance requirements rather than trust general knowledge – rules like this change. Key finding: the Attorney General charity registration clock starts at first donation received, not at incorporation – a 30-day window that’s easy to blow past if you’re not watching for it. Also confirmed the Statement of Information (SI-100) is due within 90 days of incorporation, then biennially.

Output was a 3-phase roadmap document:

  • Now – compliance filings, board recruitment, bylaws, insurance (general liability matters a lot here given power tools on-site), bookkeeping
  • Next – zoning/occupancy for the space, safety waivers, fundraising, finishing the website
  • Later – instructor recruiting, launch event, ongoing compliance calendar

Part 4: Auditing the Website

We are able to connect Claude directly to the website by using the Chrome extension. The process was simple: install the Chrome extension, log in with my own WordPress credentials myself (Claude never sees the password), and once I was in, Claude had admin access to navigate and read the site directly. Went through all 11 pages and both posts by hand.

Findings:

  • Home and About are real – actually written, not template filler
  • Contact, Volunteer, FAQ still have the starter theme’s placeholder content (fake Chicago address, Lorem Ipsum, a Golden Gate Bridge map embed, paid "science camp" events that have nothing to do with us)
  • Library page is blank
  • The 8 fake "Stories" blog posts were already trashed – by someone, at some point – but the SiteGround cache was still serving the old version publicly. Needed one click on "Purge SG Cache," not a content fix
  • Two duplicate "Donor Dashboard" pages (a GiveWP plugin quirk) – one needs trashing
  • One post, titled "Elephant and Goldfish Model – Dave Rensin", was initially misjudged during the audit as a stray pasted article and briefly trashed – it is actually legitimate project documentation. Caught the mistake, restored it, and republished.
  • This post’s predecessor, the CNC VCP guide, is real project documentation and stays

Part 5: Site Cleanup

Went back through the flagged items from the audit. Trashed the duplicate "Donor Dashboard" page. Restored and republished the Elephant and Goldfish post once we confirmed it was real. Wrote real copy for Contact, Volunteer, and FAQ – replacing the starter theme’s fake Chicago address, Lorem Ipsum, a Golden Gate Bridge map embed, and paid "science camp" events that had nothing to do with us. Kept the working Kadence contact and volunteer forms intact and only touched static text.

Contact page keeps things simple: email only, no phone or physical address for now. FAQ went from 9 placeholder Q&As to real ones covering pricing (free), our pre-lease status, tool experience requirements, how to get involved before launch, our 501(c)(3) status, and how donations work.

Still pending: the starter theme’s stock "kid science" photos are still sitting on several pages. Real photos are coming – they will get swapped in once uploaded.

Part 6: Getting Official

Added the EIN and 501(c)(3) status to the site: once in the sitewide footer, and again as a proper disclosure line on the About page. The footer edit turned into its own small debugging story – the Kadence Customizer’s copyright field ended up with garbled, duplicated text after a couple of failed manual edits in its Visual/Code tabs. Turned out the corruption was not a caching issue – the saved Customizer setting itself (footer_html_content) held the bad value. Fixed it by reading and writing that setting directly through wp.customize() in the browser console, republishing, and confirming the live footer rendered clean before moving on.

Part 7: Navigation

Added a "Stories" link to the main menu so the blog is actually reachable from the site instead of only existing as a URL.


Part 8: Real Photos, Finally

Got a batch of 30 real photos uploaded. Swapped the Contact page hero (a stock photo of kids with a microscope) for a real shot of people at the space, then iterated on the crop several times based on feedback – moving the focus point, reverting a zoomed-out letterboxed version that ended up looking too dark, and nudging it again so a third person in the shot was actually identifiable.

The FAQ page turned out to have three images hotlinked directly from the theme vendor’s own demo server – not even hosted on our site. Replaced all three with real CNC router photos.

The Stories (blog index) page background needed a different fix than a regular page: it is controlled by the theme’s Customizer settings rather than page content, since blog archive headers work differently from static pages. Swapped its stock kid photo for a real group photo there instead.

The Volunteer/"Community" page hero went through two rounds – first a music-circle photo, then swapped again per request to a guitar-and-banjo trio, cropped just above their heads.

Part 9: Menu and Navigation Cleanup

Removed a dead "Events" link that pointed to a page that never existed. The Library page has no content yet, so it is now a draft and off the main menu – easy to bring back later.

Also found the footer menu had drifted out of sync with the main menu: it still listed the now-offline Library page through a completely broken custom link to an old URL, and its "About" link was actually pointing to the FAQ page under the wrong label. Rebuilt the footer menu to exactly match the header: Home, About, Community, Stories, Contact.

Along the way, confirmed the nav’s "Community" link is not a separate page – it is the Volunteer page relabeled, which lines up with the plan to broaden that page’s content beyond just volunteering.

Part 10: Page Copy Touch-ups

Swapped this post’s own intro paragraph to better match the org’s voice: "The Fertile Soil freely shares our knowledge. With that in mind, we are sharing our workflow for today."

The Volunteer page’s intro line is now a single sentence – "Volunteering is just one of the many ways to join The Fertile Soil community." – a placeholder ahead of a full rewrite of that page to talk about community more broadly. Also flagged a broken image reference still sitting in that page’s body (an old stock photo file that no longer exists on the server) for cleanup whenever the fuller rewrite happens.


Key Lessons Learned

  1. Claude’s browser automation only works through the Claude in Chrome extension – it can’t drive Safari, only view it.
  2. A connector named after a platform (WordPress.com) doesn’t cover every site running that software – self-hosted installs need direct admin access instead.
  3. The nonprofit compliance clock starts at first donation, not incorporation. Worth building this into the calendar now, before any money comes in.
  4. Caching plugins can make already-fixed content look broken on the live site – check the CMS state directly before assuming something needs redoing.
  5. Not everything from a starter theme is junk – About and Home were already real, the placeholder blog posts were already trashed. Auditing page-by-page beats assuming the whole thing needs a rewrite.
  6. When automating rich-text or Customizer fields, verify the saved value directly through the underlying JS API or option, not just what renders in the editor – a field can look right in the UI while the stored value is already corrupted.
  7. If the site owner says a piece of content is legitimate, that overrides an editor’s own guess about what looks like a "stray" post.
  8. A menu label does not always match the page behind it – worth checking what a nav item’s underlying page actually is before assuming the label is accurate.
  9. Header and footer menus can drift apart over time; when one gets fixed, check whether the other needs the same fix.

July 11, 2026. Next up: a fuller rewrite of the Volunteer/Community page content, and getting the roadmap’s compliance deadlines onto an actual calendar.

Similar Posts