Claude Code · in Practice

Handout Directory

Real workflows for real problems. Click any entry to see what it solves, when to reach for it, and how it actually works.

Agentic workflow

Agentic development workflow in 9 steps — from brief to commit. Click any card to open the full interactive pipeline.

01AI command/feature-refinementFeature refinement → 02Human review👁 ReviewBrief skim → 03AI command/iterative-reviewBrief iterative review → 04Human review👁 ReviewFinal brief review → 05AI command/plan-makerDetailed plan → 06AI command/iterative-reviewPlan iterative review → 07Human review👁 ReviewFinal plan review → 08AI command/implement nextImplement next task → 09Human review👁 ReviewImplementation review

Skills

/feature-refinement Turn a rough idea into a brief you can hand off. A senior product thinker walks you through the questions you'd otherwise skip, and lands you with a scoped Feature Brief ready for /plan-maker. /aaa When "looks good to me" isn't enough. Stress-tests an idea, feature, or design against real-world best practices and hands you 3–4 concrete ways to make it world-class. /documentation-standard Docs your team will actually find again. Creates and refactors architecture notes, ADRs, manuals, and dev guides with enforced structure — nothing gets lost in a folder. /llm-wiki You've done the research — but Claude keeps forgetting it. Captures your notes, sources, and decisions in one place; future chats search it first → sharper answers, fewer tokens. /plan-maker Stop staring at a ticket wondering where to start. Breaks a feature into the smallest tasks with tests and dependencies — ready for /implement to chew through. /skill-packager Built a Claude Code skill? Make it work in Claude Desktop too. Packages your skill folder into an upload-ready ZIP for Desktop and claude.ai — names, paths, and frontmatter handled. /wrap-up You've been coding for hours — but is it actually done? Runs tests, devil's-advocate, and a commit audit, then tells you exactly what's left. Never commits or fixes anything without an explicit yes. /clean-code-review Code done, want the real read. Runs 132 checks across 7 groups — clarity, smells, SOLID, architecture, tests, safety, DDD — on local changes, a git range, a pull/merge-request link, or specific files. /commit Commit time, message not. Reads the staged diff and writes a Conventional Commits message with a why-first body, then commits — or drafts the text only with /commit message. /doc-voice Docs that read like marketing copy or dry internal prose. Applies a problem-first, proof-led voice to READMEs, handouts, and guides — without touching structure. /md-reviewer Terms shift, cross-references rot, contradictions accumulate. Reviews Markdown for consistency, contradictions, and stale links — handles sets of 50+ files. /options Stuck between approaches. Produces 2–4 genuinely different paths with honest pros/cons, grounded in your actual project files, and a firm recommendation. /plain-language Need to explain a technical decision to a stakeholder who doesn't know the code. Rewrites it consequence-first, technical detail in parentheses — precise but followable. /quick-plan No plan yet and too busy for /plan-maker. Defines the goal, success criteria, and 4–12 steps inline — one pass, no file written. /bugfix You have a bug and need it gone — not just patched. Drives a four-agent pipeline: failing test first, TDD fix, doc update, full review loop, commit. /status-report Kicked off a long task and don't know when it'll finish. Reports status on demand or on a recurring schedule — cancel anytime with off. /session-log Never lose a Claude conversation again. Every prompt and the answer it got are quietly saved into per-project Markdown logs — working time, estimated tokens and price, model and effort switches, and a line for every sub-agent that ran — organized by session, searchable, shareable, yours to keep.

Review & execution skills

/da-review A second opinion that actually pushes back. Runs a one-pass devil's-advocate review of your code, plan, or design and lists every flaw it can find — no auto-fixes, just honest verdicts. /implement Execute a plan, one task or the whole thing. next builds the next unchecked task test-first, reviews, and commits — one task, done; all loops every remaining task hands-off, a subagent per task wherever the harness has a subagent tool (Claude Code including claude -p, OpenCode, omp, Cursor), falling back to inline only where none exists; all inline forces inline. Works in any harness. /iterative-review A review that doesn't stop at finding problems. Multiple devil's-advocate agents stress-test your work in parallel; fix agents patch every major issue; loops until the pass is clean.

Agents

devils-advocate The agent that hunts for what can break. Stress-tests a plan, code, or recommendation by attacking assumptions and exposing blind spots — used inside /da-review and /iterative-review.

Scripts

scripts-plan See what's next without re-reading the whole plan. Prints a progress header for the next uncompleted task — /implement reads it once in Step 1 in NEXT mode, on every iteration in ALL mode.