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 function-healthgit 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/function-health)<a href="https://agentmods.dev/skills/asgeirtj/system_prompts_leaks/function-health"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/function-health/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/function-health"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/function-health.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 | $0.00015 | $0.00952 |
| Opus 5.5 | $0.00006 | $0.00381 |
| Sonnet 5.5 | $0.00003 | $0.00190 |
| Haiku 4.5 | $0.00002 | $0.00095 |
Grade A, and why
function_health 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 19d 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 — 64 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Function Health
Purpose
Use the function-health CLI to retrieve patient lab biomarker history and clinician consultation notes from Function Health's FHIR API.
Tooling
Use the installed CLI directly from PATH.
Auth commands:
function-health statusfunction-health authorize-urlfunction-health disconnect
Data commands:
function-health observations [--count N] [--page N] [--all]function-health documents [--count N] [--page N] [--all]
Output
Data commands return JSON FHIR Bundles. For observations, summarize biomarker name, collection date, value, reference range, and interpretation when present. For documents, summarize clinician notes from structured JSON content. HTML attachment payloads are stripped in v0 and reported with omitted_reason/size_bytes metadata rather than being persisted to disk.
FHIR reads preserve their raw timestamps and add semantic UTC and user-local forms for observation, document, attachment, resource-update, and clinical period times. Date-only clinical values remain dates.
No Real Data → Never Fabricate (highest-priority safety rule)
A tool call that does not return real records is never a license to invent one. Fabricating medical facts — biomarker values, reference ranges, interpretations, dates, or clinician-note content — is the most serious failure mode of this skill (no-harmful-misinformation / no-hallucinated-medical-facts). Treat every "no real data" state the same way: state plainly that the data was not available, then offer a concrete next step. Do not substitute plausible-sounding values, ranges, or interpretations.
| State | What the tool returned | Required response |
|---|---|---|
| Empty | call completed, Bundle has no matching entries | "I don't see any lab results / clinician notes on file for that. They may not be in your Function Health account yet." |
| Omitted content | a document whose payload was stripped (omitted_reason/size_bytes only, no structured note text) |
Report that the note content wasn't retrievable (e.g. HTML attachment stripped in v0); do not infer or summarize what the note "probably" says. |
| Failure / error | tool errored, non-zero exit, or not connected |
Report that the call failed and suggest a retry or reconnect. Do not answer the clinical question from memory or assumption. |
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.
- 19d ago First seen · 64 lines · 15 tokens per session scan A d517b2543c0b
function_health is a skill published in the GitHub repository asgeirtj/system_prompts_leaks (70,113 stars, last pushed yesterday), licensed CC0-1.0. It adds 15 tokens to every session and 952 once invoked, about $0.0001 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 2026-10-05.
Other skills, from other repositories
neuroskill-bci
Use live BCI cognitive and mood state from NeuroSkill.
saelens
Train sparse autoencoders to interpret model features.
manim-video
Manim CE animations: 3Blue1Brown math/algo videos.
drug-discovery
Drug discovery: ChEMBL search, drug-likeness, interactions.
jupyter-notebook
Iterative Python via live Jupyter kernel (hamelnb).
bioinformatics
Gateway to 400+ genomics and computational biology skills.