design intelligence for ai coding agents. one unified cli, 36 commands across recon, review, build, plan, ship, test, reflect, power. zero npm deps.
Skill Claude Code
Generate optimal Claude Code configuration for any project. Creates CLAUDE.md, settings.json, skills, hooks, agents, and commands tailored to your stack. Run /forge to start.
Skill Claude CodeCodex
File-level accessibility audit with concrete line locations for every finding. Eight rules: img-missing-alt, button-without-label, link-without-text, input-without-label, positive-tabindex, click-handler-on-non-interactive, autofocus, low-contrast-token-pair. Use when the user wants to audit a11y, find accessibility…
Skill Claude CodeCodex
Run /brief, /plan-eng, and /plan-design in one shot to bootstrap a complete sprint plan. All three documents land in .dragoon/plans/ pre-filled with codebase facts. Use when the user wants to spin up a full plan quickly. Defaults to dry-run; pass --apply to write all three files. Run as node…
Skill Claude CodeCodex
Snapshot codebase metrics (file count, total lines, /critique scores) for trend tracking. Subcommands capture/list/compare/show. Snapshots stored as JSON in .dragoon/benchmarks/. Use to track design quality and codebase growth over time. Run as node /.claude/skills/dragoon/skills/benchmark/scripts/benchmark.js .
Skill Claude CodeCodex
Generate a structured product brief markdown from a one-line idea, pre-filled with codebase facts (stack, type system, palette, components count) from dragoon.json. Use when the user is starting a new feature, project, or initiative and wants to think through it before coding. Output goes to…
Skill Claude CodeCodex needs its repo
Generate a curl-based canary watcher script for a deployed URL. Validates URL is http/https, status code is 100-599, interval is 5-86400 seconds. The generated script supports one-shot (default) and --watch (continuous loop) modes. Wire into cron or a CI scheduled job. Run as node…
Skill Claude CodeCodex
Write a checklist of destructive commands the user wants to think twice about (rm -rf, force push, hard reset, db drop, etc.) to .dragoon/CAREFUL.md. dragoon does NOT block commands at runtime - this is a discipline tool, a list the agent reads before risky operations. Use to set explicit guardrails for an AI agent or…
Skill Claude CodeCodex
Generate a color palette token file (CSS variables, tailwind config, or JS module) from the palette detected by /scan. Uses inferred role names (foreground, background, accent, danger, etc.) where possible, numeric fallback otherwise. Use when the user wants to formalize color tokens, extract a palette from existing…
Skill Claude CodeCodex
Generate a UI component scaffold in the codebase's actual style. Reads dragoon.json to match framework (react/vue/svelte), styling layer (tailwind/css-modules/plain), language (typescript/javascript), and design tokens (radius, spacing, palette). Defaults to dry-run preview; pass --apply to write files. Use when the…
Skill Claude CodeCodex
Score a codebase's design quality 0-100 across six categories (typography, color, spacing, motion, accessibility, consistency) plus an overall weighted score. Every score is derived from concrete metrics in dragoon.json with transparent math, no LLM, no vibes. Use when the user asks "what's the design quality of this…
Skill Claude CodeCodex
Scaffold a playwright visual regression test suite that takes per-route full-page screenshots and compares against committed baselines. Pixel diffing is done by playwright's toHaveScreenshot, not by dragoon. Use when the user wants visual regression coverage. Pass --routes to set the routes to capture. Defaults to…
Skill Claude CodeCodex
Detect drift between markdown docs (README, docs/) and the actual codebase. Catches references to npm scripts that don't exist, scoped dependencies not in package.json, file paths in backticks that don't resolve on disk, and undocumented npm scripts. Use to audit documentation hygiene before a release or after a major…
Skill Claude CodeCodex
Generate a new dragoon skill scaffold (SKILL.md, scripts/ .js, test/ .test.js, README.md) into .dragoon/forge/ / for review. dragoon does NOT auto-install forged skills. After review, copy into /.dragoon/skills/ and add an entry to bin/dragoon's COMMANDS map. Use to extend the pack with custom skills. Run as node…
Skill Claude CodeCodex
Declare directories that are off-limits to edits. Stores the list in .dragoon/freeze.json. Optionally installs a git pre-commit hook that fails the commit if any frozen path is touched. Subcommands list/add/remove/hook/check. Use when isolating a refactor, locking legacy directories, or onboarding a contributor with…
Skill Claude CodeCodex
Generate a markdown dev handoff doc with everything the next engineer needs - stack, design tokens, palette, type scale, spacing grid, motion, existing components, and conventions - all pulled from dragoon.json. Use when handing off work to another engineer or documenting the codebase for new contributors. Output goes…
Skill Codex
One-line install for dragoon. Detects supported coding agents (claude-code, codex, cursor, opencode, factory, kiro) and installs the skill pack into each. Multi-host on day one. Use when the user wants to install dragoon, mentions setup, or asks how to add dragoon to their environment. The install script is ./setup at…
Skill Claude CodeCodex
Catalog of components (PascalCase files), hooks (useFoo), pages/routes, story files, and token files in the codebase. Use to understand what already exists before adding more, or to audit a design system. Heuristic-based, no AST. Run as node /.claude/skills/dragoon/skills/inventory/scripts/inventory.js [root].
Skill Claude CodeCodex
Structured root-cause investigation template. Encodes the rule "no patches until you can reproduce, predict, and explain the bug." Forces hypothesis-test pairs, evidence collection, and a prevention plan. Use when the user reports a bug or asks to fix something they don't fully understand yet. Run as node…
Skill Claude CodeCodex
Post-merge sanity check. Verifies branch, HEAD sha, that local matches remote, and that the working tree is clean. dragoon does NOT trigger deploys (platform-specific creds out of scope). Use right before a deploy step in CI or before manual prod pushes. Exits non-zero if any check fails. Run as node…
Skill Claude CodeCodex needs its repo
Architecture overview of the codebase. Reports total file/line/byte counts, top hot files (most imported), largest files by line count, orphan files (imported by nobody), and most-used external dependencies. Pure regex on import statements, no AST. Use to understand a new codebase, find dead code, or audit dependency…
Skill Claude CodeCodex
Small persistent key-value store at .dragoon/memory.json for project-specific facts dragoon (and the user) want to remember across sessions. Subcommands list/get/set/remove/clear. Keys are lowercase ascii (.//-, max 64). Values are plain text (max 8KB, max 256 keys). Use to remember conventions, decisions, gotchas.…
Skill Claude CodeCodex
Generate motion tokens (durations + easings) using values detected by /scan, falling back to material-quality defaults when nothing was detected. Produces fast/normal/slow durations and standard/enter/exit easings. Use when the user wants to formalize motion tokens or scaffold a motion system. Defaults to dry-run…
Skill Claude CodeCodex
Static performance hints (large images, missing img dimensions, no loading=lazy, sync script tags) plus optional Lighthouse JSON report ingestion that surfaces failing audits ranked by potential ms savings. Use when the user wants to audit web performance or process a Lighthouse report. Dragoon does NOT run Lighthouse…
Skill Claude CodeCodex
Generate a structured design plan markdown that respects the constraints in dragoon.json (spacing grid, type ratio, palette, font families). Forces a screens-and-flows breakdown, accessibility plan, responsive plan, and design review checklist gated on /critique and /slop. Use after /brief or before any UI work.…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: