PenguinHarness is a local-first platform in which multiple AI agents create, evaluate, optimize, and deploy agent applications. It is for people building AI software who want agents to generate applications and improve their own behavior through skills.
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 Prism-Shadow/penguin-harness --skill a2uigit clone --depth 1 https://github.com/Prism-Shadow/penguin-harnessWrote 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/prism-shadow/penguin-harness/a2ui)<a href="https://agentmods.dev/skills/prism-shadow/penguin-harness/a2ui"><img src="https://agentmods.dev/badge/skills/prism-shadow/penguin-harness/a2ui/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/prism-shadow/penguin-harness/a2ui"><img src="https://agentmods.dev/badge/skills/prism-shadow/penguin-harness/a2ui.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 | $0.00173 | $0.04967 |
| Opus 5.5 | $0.00069 | $0.01987 |
| Sonnet 5.5 | $0.00035 | $0.00993 |
| Haiku 4.5 | $0.00017 | $0.00497 |
Grade A, and why
a2ui 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 5d 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 — 256 lines — stays where its author put it; the contents beside it link to each section on GitHub.
A2UI blocks
A reply is Markdown. Where a decision, a set of inputs, a procedure, a structure or a few readings would read better as a component than as prose, write one as a fenced block: ```a2ui holding ONE JSON object, or ```mermaid holding a diagram. The Web App renders the block with its own components and theme; the CLI and the messaging channels show the same block as readable text; nothing is lost on a surface that cannot render it. You never write HTML, scripts or styles — you name a component from the catalog and fill its fields.
Before you send a reply that carries a block, run the checker (see "Check before you send"). It is the test suite for this skill: it rejects what would not render, warns about what would be hard to read, scores the draft and prints the questions a reviewer would ask.
Before you start
This skill changes how you write replies; it needs no setup. If a message only names the skill without a task — "use a2ui", "show me the blocks" — ask what the user wants to decide, enter or understand, then answer that with the fitting block. Do not demonstrate every component at once: one block that serves the question is the demonstration.
When to use a block, and when not
Use a block when it saves the reader work:
- choice — the conversation needs ONE decision from the user and you can name the realistic options (2–7, ideally 2–5). Mark one
recommendedwhen you have a view. - form — you need SEVERAL answers at once (1–6 fields) and asking them one by one would take turns.
- steps — the user will perform a procedure by hand. One instruction per step; a
warningorcautionsits on the step it applies to and is rendered above it. - callout — one thing the reader must not miss: a warning, a caution, a tip, a note. Not for ordinary paragraphs.
- mermaid — a structure or a flow with more than three parts: a pipeline, a state machine, a sequence of calls, a data model.
- weather — the user asks about the weather somewhere. Run
scripts/weather.mjs <place>(see "Get real data"), paste the block it prints, and write one sentence before it. - clock — the user asks the time somewhere, or a meeting spans time zones: one to four zones.
- countdown — a deadline or an event at a fixed instant.
- metrics — a few readings at a glance: a system snapshot (
scripts/sysinfo.mjs), a quota or budget left, the progress of a job, figures with their change. Put them in ONEmetricsblock, never one block per number.kindsays how to read a tile:reading,used,remainingorprogress.worsesays which way is bad (defaultlowforremaining, elsehigh).warnanddangercolour the tile once the value passes them.
What ships with it
4 files 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.
- 5d ago Changed · +1 lines 48f7898f3dd3
- 6d ago Changed · +84 lines · +53 tokens per session b71b2937dbc5
- 12d ago First seen · 171 lines · 120 tokens per session scan A 4bd0e5434f0b
a2ui is a skill published in the GitHub repository Prism-Shadow/penguin-harness (2,549 stars, last pushed today), licensed Apache-2.0. It adds 173 tokens to every session and 4,967 once invoked, about $0.0007 per session on Opus 5.5. 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 2026-10-04.
Other skills, from other repositories
dashboard
Update the agent's Control Center. Use when the user says "update my Control Center", "update my dashboard", "redesign my home", or "/dashboard".
ui-craft
Design, build, redesign, or modify browser-visible web interfaces with product-quality composition, visual systems, interaction states, responsive behavior, accessibility, and browser visual QA. Use for HTML, CSS, JavaScript, React, Vue, or Svelte pages and components, landing pages, dashboards, admin tools, settings…
imagegen
Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, UI mockups, product shots, or transparent-background cutouts. Use when the result should be an image asset rather than repo-native SVG, HTML/CSS, or canvas.
frontend-design
Transform UI style requirements into production-ready frontend code with systematic design tokens, accessibility compliance, and creative execution. Use when building websites, web applications, React/Vue components, dashboards, landing pages, or any web UI requiring both design consistency and aesthetic quality.
shadcn
Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI, including chat interfaces. Provides project context, component docs, and usage examples. Applies when working with shadcn/ui, component registries, presets, --preset codes, or any project with a components.json…
ui-ux-pro-max
UI/UX design intelligence for web and mobile. Searchable local database with 84 styles, 192 color palettes, 74 font pairings, 192 product types, 98 UX guidelines, 104 icon entries, 16 GSAP motion presets, and 25 chart types across 22 stacks (React, Next.js, Vue, Nuxt, Svelte, Astro, SwiftUI, React Native, Flutter…