# The Modern WooCommerce Developer Workflow — Brian Coords (Do the Woo) ## 🗂️ Triage — annotated highlights surfaced > [!note]- Cross Reference > - **Extend only through documented hooks, never the database** — the same "build on the stable layer" instinct that governs constraining AI agents to design tokens. #xref > - **Give agents WP-CLI + browser access to self-verify** — concrete instance of the maker/checker discipline: an agent that verifies its own work before reporting done. #xref > - **AI defaults to React even when unasked** — matters for choosing PHP-only blocks vs. JS blocks when an agent writes the code. #xref ## 🧭 Bib capsule Brian Coords (WooCommerce Developer Advocate), guest on *Do the Woo* podcast hosted by James Kemp and Katie Keith, 2026. https://www.youtube.com/watch?v=Bz5_gfDmkzA ## 🧠 Author's thesis (Adler Q1+Q2) WooCommerce development in 2026 has split into two layers: a stable PHP layer (hooks, filters, Store API, settings classes) that is safe to build on indefinitely, and a fast-moving block/React/Interactivity-API layer that is still settling. The durable path is to build strictly on documented, stable extension points and treat everything else — direct database access, Experimental-tier feature flags, unreviewed AI output — as provisional. AI coding agents (given WP-CLI, browser self-verification, WooCommerce's own agent skills, and a constrained block theme) speed up the mechanical work, but human judgment stays load-bearing for taste, risk, and anything payments-adjacent. ## 🧩 Key arguments - The single most repeated rule: extend through documented functions/filters/hooks/blocks/Store API; touching the database or hand-editing core breaks safe upgrades (and is unaudited attack surface). - Feature-flag tiering is a hard build gate — anything still "Experimental" is alpha and can change or be scrapped (the Product Editor is the cautionary precedent); the Beta Tester plugin surfaces the flags. - AI changes the mechanics, not the discipline: agents self-verify via WP-CLI + browser, WooCommerce's in-repo agent skills teach platform-specific quirks, and a defined block theme constrains design output — but a human still owns visual quality and payments risk. ## 🗒️ Notes — Literature atoms ### Extend WooCommerce only through documented hooks, never the database > [!quote] > "If you find yourself writing, you know, actual database level code, like you're writing SQL commands, you're probably going down the wrong track." - **Where**: ~00:15:26 — Brian Coords, on extension best practice - **Why**: Direct DB/core edits break safe upgrades; the core discipline of the whole episode. ### AI coding agents default to React even when not asked to > [!quote] > "AI loves to write React. AI loves React. It writes it very well. It will use it if you don't ask it to." - **Where**: ~00:30:04 — comparing PHP-only blocks against React/JS blocks - **Why**: Shapes the PHP-vs-React block choice when an agent is generating the code. ### Give coding agents WP-CLI plus browser access to self-verify > [!quote] > "You have WPCLI, which is a command line tool that lets you interact with everything inside of WordPress and Woo." - **Where**: ~00:34:30 — describing his standard AI-agent setup - **Why**: Self-verification (generate test data, check output) before reporting done — maker/checker in practice. ### Trusted extensions beat vibe-coded replacements for payments code > [!quote] > "But I think this is a good argument for why people should not vibe code everything and stick to trusted extensions." - **Where**: ~00:51:48 — on paying for a maintained plugin over self-building payments-adjacent code - **Why**: A maintenance-cost argument for the extension economy over DIY replacement. ## 🔖 Permanent candidates - **Build on the stable layer, treat everything else as provisional** — When a platform exposes a documented, stable extension API alongside an evolving/experimental one, durable client work goes only through the stable layer, and anything built against the experimental layer is throwaway until it graduates. ## Links - [[AI & the Future of Knowledge Work MOC]] — the "AI does the how, humans keep the why/taste/risk" division of labor is this MOC's central thread, applied to a WooCommerce dev context. - [[DevRel MOC]] — Brian Coords' advocate role (bridging Woo engineers ↔ community) and the agent-skills-in-repo onboarding pattern are reusable DevRel practices. - [[Lit · AI tools accelerate WordPress feature development]] — same "AI speeds the mechanical work but human review stays required" theme in the broader WordPress ecosystem. - [[3. Resources/Permanent Notes/A Mature Plugin License Can Cost Less Than Rebuilding Its Integration Network]] — evidence for the payments-adjacent case where a maintained extension's integration history is worth more than a cheap generated replacement. - [[3. Resources/Permanent Notes/The Maker Should Not Grade Its Own Homework]] — contrasts browser and WP-CLI self-checks with independent review; tool access improves evidence, but the producing agent still isn't an independent checker. ## ✅ Processing checklist - [x] Claim atoms carried from deep-read analysis - [ ] Permanent candidates reviewed for promotion - [ ] Version History Diff check