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/amitpatole/agent-vision/skillnpx skills add amitpatole/agent-vision --skill skillgit clone --depth 1 https://github.com/amitpatole/agent-visionWrote 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/skills/amitpatole/agent-vision/skill)<a href="https://agentmods.dev/skills/amitpatole/agent-vision/skill"><img src="https://agentmods.dev/badge/skills/amitpatole/agent-vision/skill.svg" alt="Measured on agentmods" 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 | $0.00101 | $0.00838 |
| Opus 5 | $0.00051 | $0.00419 |
| Sonnet 5 | $0.00020 | $0.00168 |
| Haiku 4.5 | $0.00010 | $0.00084 |
Grade A, and why
agent-vision 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 4d 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 — 84 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AgentVision — see your visual output before saying it's done
You cannot see what your code renders. AgentVision can. Use it to close the loop.
When to use this
Right after you create or change anything visual, and before you tell the user it's done:
- Frontend pages/components (HTML/CSS/React/Vue/Svelte)
- Generated charts, plots, dashboards
- SVG diagrams, exported PDFs, generated images
The workflow
-
Render + analyze the artifact (use
localfor a fast, no-key structural pass; use a cloud backend for semantic critique):agentvision analyze ./path/to/artifact.html --full-page --json # or, no API key needed: agentvision check ./path/to/artifact.html --json -
Read the report. Each issue has a
kind,severity, amessage, and often abbox(DOM/CV/OCR boxes are precise; vision-model boxes are advisory). Treat theissuesarray as a to-do list. -
Fix the source to resolve each real issue (overflow, low contrast, broken images, clipped/overlapping elements, console errors, …).
-
Loop to confirm you actually fixed it and didn't regress:
agentvision loop ./path/to/artifact.html --max-iter 3The loop reports a "what changed" diff and detects when you're stuck (same issues repeating).
-
Only report success when the verdict is
pass(or remaining items are explicitly acceptedwarns).
Grade against intent (not just defects)
A defect-free artifact can still be the wrong thing. When the task had an intended result (a brief, a spec, a "make it look like X"), also grade conformance — PASS then means "matches what I was asked to build," not just "nothing broken":
agentvision conform ./artifact.html \
--brief "pricing page with three tiers and a highlighted middle plan" \
--expect 'must: a "Pro" plan card is visible'
Use --expect 'must:/should:/nice: …' (repeatable) for explicit, checkable requirements;
put exact required text in quotes (those are graded deterministically via OCR). The
report's conformance field lists each requirement as satisfied/violated/uncertain, and a
violated must fails the verdict. analyze and loop accept the same
--brief/--expect/--reference. For artifacts you generate (AI images/infographics),
close the loop on the prompt: agentvision generate --generator mypkg:make_image --brief "…" --max-iter 4.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 4d ago First seen · 84 lines · 101 tokens per session scan A 46fbd3b91f18
agent-vision is a skill published in the GitHub repository amitpatole/agent-vision (11 stars, last pushed 2mo ago), licensed MIT. It adds 101 tokens to every session and 838 once invoked, about $0.0005 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
design-system
Document and define UI design system tokens, components, and guidelines.
ui-context
读取用户通过 Little Mouse Pointer 选择的桌面窗口、UI 元素或屏幕区域上下文。.
muapi-ui-design
Generate high-fidelity UI/UX mockups for mobile and web apps using Atomic Design principles — creates wireframes and design systems via muapi.ai.
accessibility
Audit and improve web accessibility following WCAG 2.2 guidelines. Use when asked to "improve accessibility", "a11y audit", "WCAG compliance", "screen reader support", "keyboard navigation", or "make accessible".
agent-desktop
Desktop automation via native OS accessibility trees using the agent-desktop CLI. Use when an AI agent needs to observe, interact with, or automate desktop applications (click buttons, fill forms, navigate menus, read UI state, toggle checkboxes, scroll, drag, type text, take screenshots, manage windows, use…
the-loop
The user built or changed something visual — a UI, an animation, a game, a generated video — and wants it verified, or asks "why does my UI look wrong", "check that the fix actually worked", "does the animation glitch". Use this to record the running thing, critique the recording against plain-language pass criteria…