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.
npx skills add asgeirtj/system_prompts_leaks --skill animated-videogit 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/skills/asgeirtj/system_prompts_leaks/animated-video)<a href="https://agentmods.dev/skills/asgeirtj/system_prompts_leaks/animated-video"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/animated-video/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/asgeirtj/system_prompts_leaks/animated-video"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/animated-video.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00007 | $0.02245 |
| Opus 5 | $0.00003 | $0.01123 |
| Sonnet 5 | $0.00001 | $0.00449 |
| Haiku 4.5 | $0.00001 | $0.00225 |
Grade A, and why
animated-video 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 12d 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 — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Animated video
Create an animated video or motion design piece rendered as an HTML page. Build a timeline-based animation with smooth transitions. Design frame-by-frame sequences with playback controls (play/pause, scrubber). Focus on visual storytelling with the Anthropic brand palette. Export-ready at a fixed aspect ratio (16:9 or 9:16). If you need to know the position of an element (eg to move a cursor or character between elements) use refs to grab the position.
ALWAYS build on the animations_v3.jsx starter for ANY animation piece — including one hosted on a design-components page (the helmet-script + x-import structure IS the starter case, not an exemption). The only exemptions: a minor animated accent inside a larger non-animation design, or the user explicitly asking you not to use the starter. Skipping the starter silently removes the user's timeline editor — scene trims, speed changes, and video export only exist when the page builds on the engine. Do NOT load animations.jsx or animations-v2.jsx alongside it (the engines share window globals — last wins). A project that already uses an older starter keeps it; don't migrate existing animations.
START by calling copy_starter_component with kind: "animations_v3.jsx" — the continuous-composition engine: the whole animation is ONE element tree rendered from one authored-time clock, so elements move, morph, and persist across section boundaries by ordinary interpolation — nothing mounts or unmounts at a boundary. It gives you <CompositionStage>, useComposition() (→ {T, CUES}), <Shot>, <Captions>, an Easing library, and interpolate() / animate() tweens. Read the file after copying.
THE AUTHORING CONTRACT (it's in the file's usage block; follow it exactly): declare the scene list as a JSON string literal in a plain inline <script> of the MAIN document — <script>window.OM_SCENES = '[{"name":"Opening","dur":3},…]';</script> (exact JSON.stringify formatting, no spaces) — NOT in a text/babel script and NOT in a sibling .jsx (only vanilla inline script literals are addressable for the editor's write-back); declare window.OM_PLAYBACK the same way; pass both through untouched as <CompositionStage scenes={window.OM_SCENES} playback={window.OM_PLAYBACK}> wrapping ONE component — the whole piece. The scene list is the user-control view (names, order, playback durations); the engine derives the cue table from it, so the literal is the single source of structure. The user edits timing on the host timeline — trim a scene's edge, or set a section's speed — and every edit writes back into your literal and reflows the composition live.
CUE-FIRST DISCIPLINE (this is what makes a piece read as one continuous video):
- Write the OM_SCENES literal FIRST — it is the piece's outline. Get {T, CUES} from
useComposition()and key ALL choreography to T and CUES.SectionName (authored seconds) — never to your own clock, never to wall-clock time. - One helper component per section for readability, but ALL of them render ALL the time inside the one tree, keyed to cues — never conditionally mounted per section. A shared element crossing a boundary is just motion whose start and end straddle a cue.
- Define exactly three motion helpers up front (e.g.
MOTION = {enter, draw, pop}wrapping Easing curves) and use no easing or transform outside them; one caption element, one visible at a time (<Captions>has this built in).
TIMING IS USER-EDITABLE (time-stretch): when the user trims or speeds a section, the engine replays that section's SAME authored slice over the new playback length — choreography keyed to T retimes, never cuts off. Authored hard cuts are content now, not structure: wrap a shot's elements in <Shot from={CUES.X} to={CUES.Y}> (visibility flips at the cues; children stay mounted so images and videos hold their readiness). A looping piece shows its last authored frame immediately before its first — make them match.
Give every motion project a tweaks panel (kind: "tweaks_panel.jsx") whose TWEAK_DEFAULTS include "motionEditor": true with a <TweakToggle label="Motion editor"> wired to it. That key is the host timeline editor's visibility gate: the user flips it in the Tweaks panel to hide or show the editor bar, and the animation, its timing data, and export are untouched either way. Declare the TWEAK_DEFAULTS literal in a plain inline <script> of the MAIN document (the /EDITMODE-BEGIN/ convention), so the flip persists.
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.
- 12d ago First seen · 56 lines · 7 tokens per session scan A e0ef2f9fce4f
animated-video is a skill published in the GitHub repository asgeirtj/system_prompts_leaks (64,614 stars, last pushed yesterday), licensed CC0-1.0. It adds 7 tokens to every session and 2,245 once invoked, about $0.0000 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
p5js
Use when users request: p5.js sketches, creative coding, generative art, interactive visualizations, canvas animations, browser-based visual art, data viz, shader effects, or any p5.js project.
audiocraft-audio-generation
AudioCraft: MusicGen text-to-music, AudioGen text-to-sound.
baoyu-comic
A creator for educational comics, including biographies and tutorials, that turns supplied content or topics into illustrated comic pages.
concept-diagrams
Generate flat, minimal educational SVG visuals as HTML.
hyperframes
Render MP4/WebM videos from HTML compositions.
touchdesigner-mcp
Control TouchDesigner via twozero MCP.