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 skills add AkshitIreddy/agent-skills --skill visual-verificationgit clone --depth 1 https://github.com/AkshitIreddy/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/skills/akshitireddy/agent-skills/visual-verification)<a href="https://agentmods.dev/skills/akshitireddy/agent-skills/visual-verification"><img src="https://agentmods.dev/badge/skills/akshitireddy/agent-skills/visual-verification/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/skills/akshitireddy/agent-skills/visual-verification"><img src="https://agentmods.dev/badge/skills/akshitireddy/agent-skills/visual-verification.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 | $0.00073 | $0.01566 |
| Opus 5 | $0.00036 | $0.00783 |
| Sonnet 5 | $0.00015 | $0.00313 |
| Haiku 4.5 | $0.00007 | $0.00157 |
Grade A, and why
visual-verification 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 8d 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 — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Visual verification
Code that compiles is not UI that works. You have not verified a visual change until you have looked at the rendered pixels. Reading the diff, passing typecheck, and green unit tests prove none of: element visible, correctly positioned, unclipped, legible contrast, animation actually moving, art not black boxes.
The rule
Before saying a visual change is done, working, fixed, or good:
- Render it — dev server + headless browser, or the app itself.
- Capture a screenshot of the specific surface and state you changed.
- For a whole-screen visual audit, also capture 4–6 overlapping close-ups that cover the screen at readable scale.
- Read every image with the Read tool and describe what you actually see.
- Compare against intent. If it does not match, iterate — do not report success.
If you cannot render it, say so plainly instead of implying it was checked.
Close-ups, then full frame
A full-frame screenshot proves composition but can hide small typography, padding, clipping, texture, and control-state defects. Whenever the task is to judge a complete screen rather than one isolated component, inspect detail before letting the overall composition influence the review:
- Divide the visible screen into 4–6 overlapping regions that collectively cover it. Prefer semantic regions such as masthead, primary copy, media, controls, and navigation over arbitrary equal tiles when the layout allows.
- Capture each region as an element screenshot or lossless crop at a scale where its real text and edges are readable.
- Inspect every close-up individually in a strict per-image loop: open one image, analyze what is visible, record any finding or a deliberate no-defect observation, and only then open the next image. A contact sheet is useful for routing, but it does not replace opening and analyzing each constituent close-up. Do not defer analysis until after a batch of images.
- Only after every close-up has been inspected, inspect the full frame for hierarchy, balance, and relationships between regions.
- After a material correction, recapture and inspect the affected close-ups first, then inspect the updated full frame so local craft and global composition are both rechecked.
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.
- 8d ago First seen · 117 lines · 73 tokens per session scan A 8d8ae0a12546
visual-verification is a skill published in the GitHub repository AkshitIreddy/agent-skills (1 stars, last pushed 11d ago), licensed MIT. It adds 73 tokens to every session and 1,566 once invoked, about $0.0004 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 2026-08-31.
Other skills, from other repositories
lov-maintain-partners
Maintain the Skill Publisher website's partners section AND align partner logo rows on event posters / hero strips: reuse lov-find-logo for brand logo discovery, normalize collected logos to a 240px-tall content canvas (retina-ready), rasterize SVGs via rsvg-convert before normalizing (so SVG viewBox padding gets…
lov-oh-my-landingpage
A landing-page creation and redesign skill that derives a product’s positioning, story, and visual direction from its actual product and brand facts. It then implements the page responsively and checks the result against the intended product message.
lov-frontend-design
A frontend design guide for turning requirements, existing screens, and real media into usable pages and components. It covers visual design, navigation, interaction details, animation, images, audio, and video.
lov-install-shadcn-ui
A setup tool for shadcn/ui, a collection of editable interface components for web projects. It installs only needed pieces and adapts their styles to the project’s existing theme and Tailwind version.
lov-install-web-logo
A website branding tool that connects an approved logo to browser icons, the web app manifest, page metadata, and related components. A favicon is the small icon shown in a browser tab; a PWA manifest describes an installable web app.
lov-kill-emoji
A UI cleanup tool that replaces unsuitable emoji in visible interface elements with matching icons or short text. It checks the surrounding meaning before changing anything.