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 wearables-commsgit 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/wearables-comms)<a href="https://agentmods.dev/skills/asgeirtj/system_prompts_leaks/wearables-comms"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/wearables-comms/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/wearables-comms"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/wearables-comms.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 | $0.00045 | $0.02907 |
| Opus 5.5 | $0.00018 | $0.01163 |
| Sonnet 5.5 | $0.00009 | $0.00581 |
| Haiku 4.5 | $0.00005 | $0.00291 |
Grade A, and why
wearables_comms 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 — 265 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Wearables Calls and Messages
Purpose
Handle call and text-message requests that originate on a wearable. Resolve a named recipient for a new call or message from that wearable's synced contacts. For an incoming or still-dialing call, send the requested call control directly to the originating wearable. Do not silently hand an action to a paired phone.
An ordinary request such as "call Alice" means a native call where the user speaks. This workflow does not cover a request for the assistant to conduct the conversation itself.
Emergency calls
If the user asks to call the 911 emergency line, including "nine one one" or a formatted variant, do not search contacts or invoke any call command. Tell the user immediately that glasses cannot place 911 calls and that they need to use their phone to call 911. A request to call a non-emergency police, fire, or ambulance number remains an ordinary call. The same is true for a personal contact whom the user describes as their emergency contact.
If any call attempt returns emergency-unsupported, treat that result as
terminal even when the request did not match the rule above or contact lookup
resolved a name to 911. Report the restriction immediately. Do not perform
another lookup or retry through another advertised calling command or device.
Default workflow: use the current wearable
For every native call or message action in this skill, use the originating
glasses' device id from the current user turn's [device_id=...] tag. This
tag is already the exact paired-device id. Do not call device.list or
device.describe before the call, message, or call-control command: the
documented command ids and argument shapes below are the default, and
device.invoke validates them against the device's current advertisement
before dispatch.
Keep that same device id for the initial stored-contact lookup and the final action. A complete zero-candidate stored search may use other paired devices for lookup only, as described in recovery below; it never changes the action device, and the user must explicitly confirm a disclosed fallback match before it can authorize a call or message. Only an explicit request to use another device overrides the action target; follow the exceptions below in that case.
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 · 265 lines · 45 tokens per session scan A 22b3c5f7f2f2
wearables_comms is a skill published in the GitHub repository asgeirtj/system_prompts_leaks (69,759 stars, last pushed yesterday), licensed CC0-1.0. It adds 45 tokens to every session and 2,907 once invoked, about $0.0002 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.
findmy
Track Apple devices/AirTags via FindMy.app on macOS.
openhue
Control Philips Hue lights, scenes, rooms via OpenHue CLI.
actual-setup
Set up Actual Computer (actual.inc) inference in Hermes.
openhue
Control Philips Hue lights and scenes via the OpenHue CLI.
camsnap
Capture frames or clips from RTSP/ONVIF cameras.