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.
npx agentmods add skills/veryfront/veryfront-code/flywheelnpx skills add veryfront/veryfront-code --skill flywheelgit clone --depth 1 https://github.com/veryfront/veryfront-codeWhat 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 | $0.00026 | $0.01475 |
| Opus 5 | $0.00013 | $0.00737 |
| Sonnet 5 | $0.00005 | $0.00295 |
| Haiku 4.5 | $0.00003 | $0.00147 |
Grade A, and why
flywheel 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 yesterday.
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 — 233 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Development Flywheel
Self-reinforcing development loop: Run → Observe → Fix → Verify → Repeat
┌─────────┐ ┌─────────┐ ┌─────────┐
│ RUN │────────▶│ OBSERVE │────────▶│ FIX │
└─────────┘ └─────────┘ └─────────┘
▲ │
│ ┌─────────┐ │
└──────────────│ VERIFY │◀─────────────┘
└─────────┘
Quick Start
# 1. Start server in background
Bash: deno task start &
# 2. Wait for ready
vf_wait_for_ready({ port: 8080 })
# 3. Open browser (Chrome MCP)
tabs_context_mcp() → tabs_create_mcp() → navigate({ url: "http://localhost:8080" })
# 4. Take baseline screenshot
computer({ action: "screenshot" })
# 5. Begin flywheel loop
Tools
Flywheel Core
| Tool | Purpose |
|---|---|
vf_wait_for_ready |
Poll until server accepts requests |
vf_get_flywheel_status |
Aggregated view: server + errors + logs + HMR |
vf_trigger_hmr |
Force browser refresh after code changes |
Server Errors & Logs
| Tool | Purpose |
|---|---|
vf_get_errors |
Compile, runtime, bundle errors |
vf_get_logs |
Server logs with filtering |
vf_clear_errors |
Clear error collector after fixing |
Browser (Chrome MCP)
| Tool | Purpose |
|---|---|
computer({ action: "screenshot" }) |
Visual verification |
read_console_messages |
Browser console errors |
read_network_requests |
Failed API calls |
Observe Loop
// Check all error sources in one call
const status = await vf_get_flywheel_status({ port: 8080 });
if (!status.server.running) {
// Server crashed - restart
}
if (status.errors.total > 0) {
// Compilation or runtime errors
// Get details: vf_get_errors()
}
if (status.logs.errors > 0) {
// Server logged errors
// Get details: vf_get_logs({ level: "error" })
}
// Browser side (Chrome MCP)
read_console_messages({ tabId, pattern: "error|Error" });
read_network_requests({ tabId, urlPattern: "/api/" });
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.
- yesterday First seen · 233 lines · 26 tokens per session scan A da9b586f7236
flywheel is a skill published in the GitHub repository veryfront/veryfront-code (18 stars, last pushed yesterday), licensed Apache-2.0. It adds 26 tokens to every session and 1,475 once invoked, about $0.0001 per session on Opus 5. 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 skills, from other repositories
recent-release
Render Rove's recent releases as a self-contained, Rove-styled HTML page (Chinese by default). Reads shipped notes from packages/kobe/CHANGELOG.md + git tags, filters to user-facing changes, and emits one standalone .html file in the claude theme (terracotta on near-black, JetBrains Mono, terminal chrome). Use when…
nifra-web
Use when building the full-stack side of a Nifra app - file routes, layouts, loaders and actions, forms, server functions, streaming SSR, islands, SSG and ISR, the query cache, and the React, Vue, Solid, Svelte, or Preact render adapters. Covers the client/server boundary rules and the hydration errors that boundary…
storybook
Storybook is the fidelity oracle, not the runtime. The converter bundles the package's compiled dist/ into dsbundle.js — the same bundle the claude.ai/design agent builds with — and generates each preview by compiling the story source module itself (hooks, fixtures, local helpers — the whole closure comes along), with…
claude-in-chrome
Use when automating web browsing, page interactions, or DevTools via Chrome browser.
3d-object
Use when modeling 3D objects, interactive scenes, or inspection models with Three.js.
analyze-performance-traces
Analyze Chrome, Chromium, Electron, React DevTools, or Perfetto-compatible JSON traces and audit user-reported profiling findings without loading large artifacts into context; prove trigger-to-render/layout chains, separate measured facts from source inference, find exact code choke points, classify forced layout and…