Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/hamr0/agentic-toolkitnpx agentmods add commands/hamr0/agentic-toolkit/live-canvasWrote 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/hamr0/agentic-toolkit/live-canvas)<a href="https://agentmods.dev/commands/hamr0/agentic-toolkit/live-canvas"><img src="https://agentmods.dev/badge/commands/hamr0/agentic-toolkit/live-canvas/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/hamr0/agentic-toolkit/live-canvas"><img src="https://agentmods.dev/badge/commands/hamr0/agentic-toolkit/live-canvas.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 | $0.00046 | $0.04905 |
| Opus 5.5 | $0.00018 | $0.01962 |
| Sonnet 5.5 | $0.00009 | $0.00981 |
| Haiku 4.5 | $0.00005 | $0.00490 |
Grade A, and why
live-canvas 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 26d 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.
This is a copy
100% identical to live-canvas — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 310 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Live Canvas Skill
Interview, generate variants, collect feedback, refine, preview, finalize. Why each rule exists: docs/product/live-canvas-README.md.
Rules
- Always ask the feedback mode with
AskUserQuestion(Live vs JSON), every run. Never auto-select or auto-detect it. - Never run the install or relaunch commands yourself. Print them; the user runs them.
- Never start the dev server, check ports, open a browser or wait for one.
- The overlay is always wired: copy
overlay-vanilla.js, callLiveCanvas.initonce, putdata-varianton every variant container. Never generate a lab without it. - Only delete what this skill created, inside
.claude-design/or listed inroutes:/overlay:. Never a user-authored file; verify each path first. - Temp files never outlive the run. On finish and on abort, clean up (Phase 8.1).
- Never use generic or predefined visual styles; infer them from the project (Phase 0).
Feedback mode
Tool names (plugin-namespaced): mcp__plugin_live-canvas-channel_live-canvas__channel_open, mcp__plugin_live-canvas-channel_live-canvas__batch_open, mcp__plugin_live-canvas-channel_live-canvas__channel_close. Every tool returns JSON {status, port, message?, ...}.
Host: run test "$CLAUDECODE" = 1. If not Claude Code (Droid, Amp, Opencode), skip the question, announce 📝 JSON mode (Live channel requires Claude Code), omit channelUrl and batchEndpoint in the overlay init, never mention the channel plugin, and go to Phase 0.
Claude Code: ask with AskUserQuestion:
| Header | Question | Options |
|---|---|---|
| Feedback mode | Pick a feedback mode | Live channel — overlay streams each Save into this session; requires a session started with live-claude, else Live refuses to start. JSON file — overlay writes feedback to .claude-design/feedback.jsonl; tell me "done" when ready; works in any session. |
If JSON: call mcp__plugin_live-canvas-channel_live-canvas__batch_open.
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.
- 26d ago Changed · -776 lines 5ca982bd3d18
- 1mo ago Changed · +2 lines · -4 tokens per session c2c02b0ed754
- 1mo ago First seen · 1,084 lines · 50 tokens per session scan A 94eb6866fb35
live-canvas is a command published in the GitHub repository hamr0/agentic-toolkit (23 stars, last pushed 7d ago), licensed Apache-2.0. It adds 46 tokens to every session and 4,905 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. It is 100% identical to live-canvas, differing in 0 lines, and is treated as a copy.
Other commands, from other repositories
live-canvas
Conduct design interviews, generate UI variations, and collect live click-to-annotate feedback that streams into the session so edits land without leaving the browser. Use when the user wants rapid iterative UI refinement, not just batched feedback.
screenshot
Screenshot a URL or document and read it visually.
scene
Summarize the contents of the current Rhino document.
audit-accessibility
Audit web page or component for WCAG accessibility compliance.
gan-design
Run a generator/evaluator design loop for frontend or visual work with bounded iterations and scoring.
annotate
Open the drawing toolbar on the page in the Playwright browser to give Claude visual feedback.