Claude Code Guide
A 9-stage flow that ties gstack, ECC, and Auto Agent together end-to-end.
The 9-stage flow that ties three companion tools together — gstack commands, ECC Skills, and Auto Agent — from idea to ship. Each stage lists only the must-run commands; click into the commands library for full usage details.
/using-superpowers→
/autoplan→
/ui-ux-pro-max→
/tdd-workflow→
/cso→
/ship
This is the master switch that turns on every ECC Skill so the rest of the commands can auto-detect and self-fire. Skip it and a lot of skills simply won't trigger.
/using-superpowersTake the CEO's view across the competitive landscape and reference implementations, then use ECC to dig up existing libraries, templates, and projects so you don't reinvent anything.
/browse [competitor URL or name] · /search-first [keywords]/office-hours uses six pointed questions to force you to clarify the goal. /autoplan runs CEO + engineering + design reviews back-to-back and produces a prioritized implementation plan. /executing-plans loads the plan and walks through it step by step.
/design-consultation clarifies the product's positioning. /design-shotgun diverges into 3–5 options for comparison. /ui-ux-pro-max picks a final style from a library of 67 styles, 96 palettes, and 56 fonts. /design-html renders an HTML preview. /frontend-design turns it into production code.
/tdd-workflow enforces the rhythm: failing test first, smallest implementation that makes it pass, then refactor — coverage stays above 80%. Once you've written code, Claude auto-fires code-reviewer for quality, security-reviewer for security, and build-error-resolver for build breakage.
When you hit a bug, run /systematic-debugging first — hypothesis → experiment → verification — instead of patching blindly. After the fix, /verification-loop runs the full sweep: tests, build, E2E. Then /verify gives the engineer-view deep check on edge cases.
/cso runs OWASP Top 10 plus STRIDE threat modeling and attaches an attack scenario to each finding. /security-review is a deeper security-only pass. /review scores quality from the engineer's view. /requesting-code-review spawns an independent subagent for an objective third-party review.
Hand /qa a staging URL and it opens a real browser, clicks through the flows, finds bugs, fixes them, and adds regression tests. /benchmark runs perf benchmarks to surface bottlenecks.
/ship syncs main, runs the tests, pushes, and opens a PR. /finishing-a-development-branch walks you through the post-test wrap-up and merge. /land-and-deploy merges and triggers CI/CD. For risky changes, gate it behind /canary first.
Before every push, jot the current progress into PROJECT_STATE.md so the next conversation picks up exactly where this one left off. Run /retro to look back at this round of work and flag the friction points so you can sidestep them next time.
Vexilo is the missing index for Claude Code — 31 agents, 125 commands, 123 skills and more, organized by scenario. Browse free; unlock search, filters and one-click CLAUDE.md export with a one-time $19 licence.