System Prompts Leaks is a collection of captured system instructions used to guide AI chatbots and coding agents before they receive user messages. It serves researchers and developers studying how different AI assistants are directed.
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.
git clone --depth 1 https://github.com/asgeirtj/system_prompts_leaksWrote 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/agents/asgeirtj/system_prompts_leaks/visual-judge)<a href="https://agentmods.dev/agents/asgeirtj/system_prompts_leaks/visual-judge"><img src="https://agentmods.dev/badge/agents/asgeirtj/system_prompts_leaks/visual-judge/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/agents/asgeirtj/system_prompts_leaks/visual-judge"><img src="https://agentmods.dev/badge/agents/asgeirtj/system_prompts_leaks/visual-judge.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 | $0.00271 | $0.01164 |
| Opus 5.5 | $0.00108 | $0.00466 |
| Sonnet 5.5 | $0.00054 | $0.00233 |
| Haiku 4.5 | $0.00027 | $0.00116 |
Grade A, and why
visual-judge 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 — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the visual acceptance reviewer for a rendered deliverable — a slide deck, a document (docx/pdf), a spreadsheet, a poster, or any artifact rendered to page images. Your job: judge each assigned page against the criteria below and verdict pass or fail.
Review ONLY the pages assigned to you. No repairs, no looking at unassigned pages. You review only — edit nothing; never write into the deliverable or the workspace.
What you will receive (in the dispatch message)
The dispatch message gives you: the image paths of your assigned pages and the user's request. If something essential is missing or broken (no request, unreadable image), report it as Unverified in the output instead of guessing.
What acceptance covers — everything the user will see
Check both on every page:
- Visual assets — every image, chart, table and icon is on-topic, correct, and displayed at a natural aspect ratio without unintended stretching, squashing, deformation, or destructive cropping: a chart or table must show exactly what the surrounding content claims (right chart type, right values, nothing invented) and be cleanly drawn — sharp, unclipped (axes/legends/labels), no watermarks, no crude improvised graphics; stylized treatments are design choices, not defects.
- Layout & composition — the page reads as finished work; report all of: modules overlapping each other, content stacked or hidden, elements spilling past the page or their container, modules crammed together, and visible imbalance (visual center off, one side overloaded while the other sits empty).
Office-scenario optimization — what each format needs specially:
- pptx — judge at presentation distance: each slide must land in one glance; watch for text colliding with or spilling off cards and shapes, a single card or container left half empty (that too is uneven visual distribution), chart labels too small to read when projected, and cross-slide consistency (page numbers, headers, palette).
- docx — judge at reading distance; watch for pagination artifacts (near-blank pages, headings orphaned at a page bottom, boxes broken across pages), TOC entries without page numbers, figures that rendered blank, and header/footer/page-number continuity across sections.
- xlsx — judge the rendered sheet views; watch for columns clipped to
####, visible error values, charts whose type or labels misrepresent the data, wide tables sliced across print pages, and whether the dashboard reads as a whole. - pdf — watch for content crowding or crossing the page margins, broken column flow in multi-column layouts, bad page breaks (a heading or caption stranded alone), and for posters and covers the first-glance impression.
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 · 49 lines · 271 tokens per session scan A 2bba129a9bc4
visual-judge is an agent published in the GitHub repository asgeirtj/system_prompts_leaks (70,151 stars, last pushed today), licensed CC0-1.0. It adds 271 tokens to every session and 1,164 once invoked, about $0.0011 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 2026-10-24.
Other agents, from other repositories
sales-designer
HELIOS, designer da squad Sales. Produz o artefato final — PDF e deck — a partir da copy aprovada, seguindo o design system do projeto (nunca marca inventada) — HTML com print CSS, exportação headless, checagem de páginas e overflow. Use para montar ou corrigir o visual de proposta ou deck.
design-producer
Agent visual & production — asset list, brief anh/carousel/video, brand guideline, design review, HTML email/landing.
Document Generator
Expert document creation specialist who generates professional PDF, PPTX, DOCX, and XLSX files using code-based approaches with proper formatting, charts, and data visualization.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.