DS Skills Pipeline is a command-line workflow that turns a design system's source code into structured skills for coding agents. It extracts verified tokens, components, icons, assets, usage patterns, and import paths, then generates and verifies reference files that agents can use when working with the design system.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/vercel-labs/design-systems-to-agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/vercel-labs/design-systems-to-agent-skills/4-generate)<a href="https://agentmods.dev/commands/vercel-labs/design-systems-to-agent-skills/4-generate"><img src="https://agentmods.dev/badge/commands/vercel-labs/design-systems-to-agent-skills/4-generate/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/vercel-labs/design-systems-to-agent-skills/4-generate"><img src="https://agentmods.dev/badge/commands/vercel-labs/design-systems-to-agent-skills/4-generate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.03251 |
| Opus 5 | $0.00000 | $0.01625 |
| Sonnet 5 | $0.00000 | $0.00650 |
| Haiku 4.5 | $0.00000 | $0.00325 |
Grade A, and why
4-generate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 328 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Stage 4: Parallel Skill File Generation
Required Capabilities
- File read/write
- Shell command execution (git, directory creation)
- Sub-agent spawning (recommended for parallelism; can dispatch serially without it)
Objective
Run Stage 4 of the design system skill generation pipeline: parallel skill file generation.
This should run in a fresh agent session. If this session has significant conversation history from Stages 1-3, start a new session first.
Read the closed PRD, decisions, and verified facts from disk. Generate all skill files using parallel sub-agents in batches of 8. Each component batch runs in its own fresh session to avoid context accumulation — the progress file on disk provides continuity between sessions.
The PRD may be either a single file (03-closed-prd.md) or a directory of 5 files (03-closed-prd/). Both formats are supported — detect which is present and load accordingly.
Process
Step 0: Session check
Check if this session has existing conversation history beyond loading this command.
If yes, warn:
This session has existing context. Stage 4 is the heaviest stage — it generates all skill files and needs maximum context budget.
Recommended: Start a fresh agent session and run Stage 4 again.
Continue anyway? (The mandatory pauses and progress file will help manage context, but fresh is better.)
Step 1: Load context from disk
Identify the design system by scanning context/ for available directories.
PRD format detection
Check which PRD format is present:
- Directory format (
context/{ds}/03-closed-prd/exists): Multi-file mode. Load files selectively based on the current phase (see table below). - Single-file format (
context/{ds}/03-closed-prd.mdexists): Legacy mode. Load the full file.
Selective PRD loading (directory format only):
| Phase | PRD files loaded |
|---|---|
| Wave 1 (infrastructure) | 01-file-manifest.md + 02-content-structure.md + 03-wave-plan.md |
| Wave 2 (guides) | 02-content-structure.md + 03-wave-plan.md |
| Wave 3+ (component batches) | 05-subagent-template.md + 03-wave-plan.md |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 328 lines · 0 tokens per session scan A 05413eb61fcd
4-generate is a command published in the GitHub repository vercel-labs/design-systems-to-agent-skills (37 stars, last pushed 4mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,251 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
profile
Save, load, inspect, update, reset, or delete diagram-design client profiles.
ui-flow-review
Review menus, HUD, navigation, and player flow from a UX perspective.
responsive-design-specialist
Use when a layout breaks between sizes. Arbitrary breakpoints, type that does not scale, images that blow out the grid, or a desktop design retrofitted onto mobile.
ppt-image2-editable-rebuild
Rebuild image2 or imagegen reference slides as editable PowerPoint decks.
p3-ux-wireframes
Creates wireframes (ASCII art) for the most important screens with interaction descriptions.
team-ui
Orchestrate UI/UX design team.