Simplicity-First × AI

AI Is an Amplifier, Not a Simplifier.

Without constraints, AI helps teams build complexity faster. With the right constraints, AI becomes the most powerful simplicity tool a team has ever had. The four skills below are those constraints — ready-to-use Claude skill files that bake the Simplicity-First philosophy into how your AI assistant designs, reviews, and audits code.

Browse the Skills ↓ How Skills Work

Why Simplicity-First Matters More When AI Is Writing Code

AI assistants default to comprehensive solutions. Comprehensive means complex. They reach for microservices when a module would do, suggest plugin systems when one implementation exists, and generate elaborate abstractions for problems that don't yet exist.

They also ship faster than human review can absorb. Working code lands in pull requests before anyone has formed an opinion about whether the approach is right. Cognitive debt — code the team does not understand well enough to change safely — compounds silently.

The skills on this page are the constraint set that flips that default. They tell your AI assistant to design for the 2 AM engineer, build half of what it thinks you need, and refuse to merge what the author can't explain.

How Claude Skills Work

A skill is a packaged set of instructions that activates automatically when Claude detects the right kind of request. You install it once. It applies on every relevant task.

Download

Each skill is a single .skill file. Pick the ones that match the work your team does — architecture design, code review, PR readiness, codebase audits.

Install

Install the skill in Claude (or any AI assistant that supports the skill format). Each skill includes a SKILL.md file with the full instruction set.

Use

Ask normally. The skill activates when your request matches its triggers — you don't need to invoke it by name. The output comes back filtered through Simplicity-First.

The Simplicity-First Skills Bundle

Four skills. Four moments where AI tends to add complexity. Four constraints that turn AI into a simplicity tool instead.

simplicity-first-architect.skill

Simplicity-First Architect

Applies the Simplicity-First philosophy to software architecture work. Goal: designs real teams can actually maintain at 2 AM, not designs that impress at a review board. Built around the three filters — the 2 AM Test, the Half-Rule, and Primary Path First.

When it activates

› “Design a system for…”
› “Microservices or a monolith?”
› “Review this architecture”
› “What pattern should I use for…”

What you get back

› The design, lead with the primary path
› What was considered and rejected, with reasons
› A walkthrough of how the design holds up at 2 AM
› Named seams for future expansion — not implemented

Download simplicity-first-architect.skill

simplicity-first-reviewer.skill

Simplicity-First Reviewer

Reviews code changes through the Simplicity-First lens to catch unnecessary complexity before it ships. Generic AI reviewers catch style; this one catches what will hurt the team at 2 AM. Runs four checks: complexity additions, tribal-knowledge dependence, 2 AM impact, and whether a simpler alternative exists.

When it activates

› “Review this PR / diff / patch”
› “Should we merge this?”
› “Did I miss anything in this code?”
› A diff or patch pasted with no instructions

What you get back

› A direct verdict: merge, revise, or reconsider
› PASS / CONCERN / FAIL on each of the four checks
› Concrete simpler alternatives, not vague hedges
› Stricter checks on AI-generated code, not looser

Download simplicity-first-reviewer.skill

explain-or-dont-merge.skill

Explain or Don't Merge

Prevents cognitive debt at the place where it accumulates: pull request time. The rule is simple. If the author can't explain the change in their own words, the change does not merge. Whether they wrote it themselves or had Copilot, Cursor, or Claude Code generate it.

When it activates

› “Help me explain this PR”
› “I had Claude write this. Should I merge it?”
› “Is this ready to merge?”
› Generated code pasted with “OK to ship?”

What you get back

› Four questions to walk through in your own words
› An honest evaluation of whether the explanation matches the code
› A ready-to-paste PR description when you understand it
› A clear “not yet” with named gaps when you don't

Download explain-or-dont-merge.skill

simplicity-audit.skill

Simplicity Audit

The offensive use of AI: instead of letting it add complexity, use it to find and eliminate the complexity that's already there. Audits an existing codebase through the four 2 AM dimensions — discoverability, traceability, change safety, and cognitive load — and produces a ranked list of complexity hotspots with proposed simplifications.

When it activates

› “Audit this codebase”
› “Where is the complexity in our system?”
› “What should we refactor?”
› “Where would a new engineer get stuck?”

What you get back

› Up to five complexity hotspots, ranked by severity
› Specific files, the failure scenario, and effort estimate
› Proposed simplifications that preserve current behavior
› An honest accounting of what the audit could not see

Download simplicity-audit.skill

Use AI to Subtract.

Every line AI helps you delete is energy you no longer consume, code you no longer maintain, and one less thing for a tired engineer to get lost in at 2 AM.

Read the Manifesto → Architecture Assessment