1. Do more in parallel with git worktrees
The team's number one tip and the single biggest productivity unlock. Spin up 3 to 5 git worktrees at once, each running its own Claude session in parallel. Worktrees beat checkouts because the team prefers them so much that native worktree support got built into the Claude Desktop app. Some engineers name their worktrees and set shell aliases like za, zb, zc so they can hop between them in one keystroke. Others keep a dedicated "analysis" worktree just for reading logs and running BigQuery.
Marketplace context: imagine running five Claude sessions on five copies of your store backend. One refactors the checkout, one writes tests, one drafts docs, one explores a new feature, one analyzes recent order data. You stop waiting on the agent and start orchestrating a small team.
2. Start every complex task in plan mode
Pour your energy into the plan so Claude can one-shot the implementation. Don't keep pushing when things go sideways — switch back to plan mode and re-plan. One team member runs one Claude to write the plan, then spins up a second Claude to review it as a staff engineer. Another switches to plan mode the moment anything drifts. They also explicitly tell Claude to enter plan mode for verification steps, not just for the build.
Marketplace context: when you're building anything non-trivial, like a listing optimizer or a multi-marketplace inventory sync, plan mode is where data sources, edge cases and acceptance criteria get locked in before any code exists. Cuts rework dramatically.
3. Invest in your CLAUDE.md
After every correction, end with: "Update your CLAUDE.md so you don't make that mistake again." Claude is eerily good at writing rules for itself. The team ruthlessly edits CLAUDE.md over time, iterating until the mistake rate measurably drops. One engineer has Claude maintain a notes directory for every task, updated after every PR, and points CLAUDE.md at it. This is compounding engineering taken to its logical extreme — the AI trains itself on its own failures.
Marketplace context: your CLAUDE.md ends up encoding business logic. Things like "always validate SKU format before hitting the API," "Shopify rate limits are 2 req/sec," "never overwrite parent product images." The agent stops repeating the same expensive mistakes.
4. Create your own skills
Skills are reusable instructions Claude loads automatically. If you do something more than once a day, turn it into a skill or slash command. Commit them to git so the whole team inherits the workflow. Team examples: a /techdebt command that runs at the end of every session to find and kill duplicated code. A context-dump command that pulls 7 days of Slack, GDrive, Asana and GitHub into one briefing. Analytics-engineer-style agents that write dbt models, review code and test changes in dev.
Marketplace context: build a "catalog enricher" skill that processes raw product data into clean listings. A "PDP auditor" that checks every product page against brand rules. A "review responder" that drafts replies in your voice. Skills make repeated work disappear.
5. Let Claude fix most bugs by itself
Enable the Slack MCP, paste a bug thread into Claude, just say "fix." Zero context switching. Or just say "go fix the failing CI tests" and don't micromanage how. Point Claude at docker logs for distributed systems work — it's surprisingly capable.
Marketplace context: when a customer hits a checkout error in your support chat, paste the thread into Claude with the codebase open. It often finds the root cause faster than the back-and-forth between support and dev.
6. Level up your prompting
Three patterns from the team. First, challenge Claude. Say "grill me on these changes and don't make a PR until I pass your test." Make Claude be your reviewer. Or say "prove to me this works" and have it diff behavior between main and your feature branch. Second, after a mediocre fix, say "knowing everything you know now, scrap this and implement the elegant solution." Don't accept the first answer. Push for better. Third, write detailed specs. The more specific the brief, the better the output. Ambiguity at the start costs you context and quality at the end.
7. Terminal and environment setup
The team loves Ghostty for its synchronized rendering, 24-bit color and proper unicode. For easier juggling, customize your status bar to always show context usage and current git branch. Color-code and name your terminal tabs. Use tmux for one tab per worktree. The single most impactful setup tip: voice dictation. You speak 3x faster than you type, and your prompts get way more detailed as a result. On macOS, hit fn twice.
8. Use subagents
Three ways. Append "use subagents" to any request where you want Claude to throw more compute at the problem — it spawns parallel agents that explore in different directions. Offload individual tasks to subagents to keep your main agent's context window clean and focused. The main agent stays sharp because it isn't drowning in tool output from explorations. Route permission requests to Opus 4.5 via a hook. Let it scan for attacks and auto-approve the safe ones. Less interruption, same safety.
9. Use Claude for data and analytics
Ask Claude Code to use the bq CLI to pull and analyze metrics on the fly. The team has a BigQuery skill checked into the codebase. Boris's own quote: "Personally, I haven't written a line of SQL in 6+ months."
Marketplace context: This works for any database with a CLI, MCP or API. For ecom that means your Shopify, Stripe, GA4 and warehouse data all become conversational. Ask Claude "which SKUs had the worst margin in November and why" and get a real answer with the underlying query attached.
10. Use Claude Code as a learning tool
This is the most underrated tip. Enable the Explanatory or Learning output style in /config to have Claude explain the why behind every change. Ask Claude to generate a visual HTML presentation explaining unfamiliar code — it makes surprisingly good slides. Ask for ASCII diagrams of new protocols and codebases. Build a spaced-repetition learning skill where you explain your understanding, Claude asks follow-ups, and stores the result for review. Claude Code isn't just for writing code. It's a powerful learning tool when you configure it to explain and teach.
Why this matters now
These aren't hot takes. They're the actual workflow patterns used by the team that builds Claude Code itself. If you're an ecom founder or operator treating AI like fancy autocomplete, you're leaving most of the value on the table.
Parallelize with worktrees. Plan before you build. Let the system learn from its mistakes. Build skills around your repeated work. Verify everything.
That's how solo operators are starting to out-build small dev teams in 2026.
