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 agentmods add instructions/imavgerix/exocontrol/agents-mdgit clone --depth 1 https://github.com/ImAvgErix/ExoControlWrote 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/instructions/imavgerix/exocontrol/agents-md)<a href="https://agentmods.dev/instructions/imavgerix/exocontrol/agents-md"><img src="https://agentmods.dev/badge/instructions/imavgerix/exocontrol/agents-md.svg" alt="Measured on agentmods" 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 | $0.01124 | $0.01124 |
| Opus 5 | $0.00562 | $0.00562 |
| Sonnet 5 | $0.00225 | $0.00225 |
| Haiku 4.5 | $0.00112 | $0.00112 |
Grade A, and why
ExoControl AGENTS.md 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 4d 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 — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Exo Control — agent instructions (any model, any harness)
You have eyes and hands on a Windows PC via Exo Control. You are not limited to one vendor UI.
How you call it
- MCP tools (preferred if available):
exo_exec,exo_screenshot,exo_help - CLI:
exo-control exec --steps '...'orexo-control script file.json - Python:
ExoExecEngine().execute([...])(alias:AetherExecEngine)
If you do not know an op, run {"op":"help"} or exo_help first.
Non-negotiable workflow
- Lease before hands:
lease_acquirewith youragent_id+ shorttask— orsession_opento hold the desk like remote access - Plan a script — batch many steps in one
exo_execcall - Focus the window (
titlesubstring; optionalmonitor) - Observe/read structure — do not screenshot first
- Act (click/type/fill/scroll/browser_* or raw
pointer/mouse/keypress/drive) then verify - lease_release or
session_closewhen done
You are a person at the desk. Aim the pointer. Roll the wheel on the document. Glance after you move.
Do / don't
| Do | Don't |
|---|---|
| UIA / DOM / refs | Coordinate spam |
scroll / scroll_into_view / browser_scroll |
Home / End (they jump the caret, they are not scroll) |
Read seen after hands (live eyes) |
Assume the screen did not change |
require_change when UI should flip |
Assume click worked |
confirm=true for kill/registry write/delete |
Silent destructive OS ops or confirm to escape allowroots |
| Fail closed and report step errors | Invent window titles or UI text |
| Compact observe | Dump full trees / raw HTML |
Safety hard stops
No anti-cheat kill, no unnamed-PID kill, no silent elevation. lease_status does not include the token. Files stay in EXO_FILE_ROOTS. See SECURITY.md.
Exo Launcher (optional target)
Installed at %LOCALAPPDATA%\ExoLauncher\app\ExoLauncher.exe. Prefer CDP/DOM when CDP is up; UIA otherwise. Not required for Control to be useful — Notepad, browsers, and any Win app work too.
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.
- 4d ago First seen · 65 lines · 1,124 tokens per session scan A de0b37095a68
ExoControl AGENTS.md is an instructions file published in the GitHub repository ImAvgErix/ExoControl (0 stars, last pushed 19d ago), licensed MIT. It adds 1,124 tokens to every session, about $0.0056 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 2026-08-31.
Other instructions, from other repositories
Shinken CLAUDE.md
Instructions for Meirtz/Shinken, covering claude.md — guide for ai coding sessions in this repo, ⛔ the one hard rule: this is a public open-source project, layout, development commands and architecture (how the pieces connect).
agent-computer-use-mcp AGENTS.md
AGENTS.md instructions for astraclawteam/agent-computer-use-mcp, covering agent worker rules, read order, repository purpose, project structure and branch and pr rules.
windows-computer-use-mcp CLAUDE.md
Claude Code instructions for sshh12/windows-computer-use-mcp, covering claude.md, what this is, commands, architecture and critical gotchas.
helix-pilot AGENTS.md
Instructions for tsunamayo7/helix-pilot, covering agents.md — helix-pilot, shared memory (mem0), connection spec (must match across all tools), http api (for codex cli, curl, scripts) and search.
helix-pilot CLAUDE.md
Instructions for tsunamayo7/helix-pilot, covering claude.md — helix-pilot, プロジェクト概要, 技術スタック, 共有記憶（mem0） and 開発ルール.
Screenhand CLAUDE.md
Claude Code instructions for manushi4/Screenhand, covering claude.md, project, commands, architecture and key entry points.