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 Stijnman/grok-custom-skills --skill desktop-subagent-connectorgit clone --depth 1 https://github.com/Stijnman/grok-custom-skillsWrote 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/stijnman/grok-custom-skills/desktop-subagent-connector)<a href="https://agentmods.dev/skills/stijnman/grok-custom-skills/desktop-subagent-connector"><img src="https://agentmods.dev/badge/skills/stijnman/grok-custom-skills/desktop-subagent-connector/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/stijnman/grok-custom-skills/desktop-subagent-connector"><img src="https://agentmods.dev/badge/skills/stijnman/grok-custom-skills/desktop-subagent-connector.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Tool Misuse · line 94 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
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 | $0.00048 | $0.01362 |
| Opus 5 | $0.00024 | $0.00681 |
| Sonnet 5 | $0.00010 | $0.00272 |
| Haiku 4.5 | $0.00005 | $0.00136 |
Grade A, and why
desktop-subagent-connector scanned grade A with 1 finding 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -sS -H "Authorization: Bearer $DESKTOP_BRIDGE_TOKEN" "$DESKTOP_BRIDGE_URL/health" How it starts
The opening of the file, as written. The whole thing — 135 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Desktop Subagent Connector
Overview
Bridges the remote sandbox to the user's real machine. The sandbox never gets raw network access to the user's LAN. Instead the user runs a small local daemon and exposes it via an outbound tunnel (cloudflared recommended). All calls are authenticated with a shared token and confined to allowed directory roots.
This is the concrete path for:
- Reading/writing files on the user's desktop
- Running approved local commands (ffmpeg, licensed tools, etc.)
- Optional browser control on the user's machine (Playwright or alongside Kimi WebBridge)
- Geo-restricted or login-session work that the remote sandbox cannot do alone
Prerequisites (user side)
- Python 3.9+
- Optional but strongly recommended:
cloudflared(makes the public URL automatic) - Optional for browser control:
pip install playwright && playwright install chromium
Easy setup (recommended)
macOS / Linux
- Get the skill scripts folder (or at least
local-daemon.py+setup.sh) onto your machine. - Run:
cd /path/to/desktop-subagent-connector/scripts
bash setup.sh
That single command:
- creates
~/desktop-subagent - generates a strong token
- starts the local daemon
- starts
cloudflaredif installed - prints a ready-to-paste block:
DESKTOP_BRIDGE_URL=https://xxxx.trycloudflare.com
DESKTOP_BRIDGE_TOKEN=...
Paste that block back to the agent. Leave the terminal open (Ctrl+C stops everything).
If cloudflared is missing:
# macOS
brew install cloudflared
# then re-run setup.sh
Windows (PowerShell)
cd path\to\desktop-subagent-connector\scripts
powershell -ExecutionPolicy Bypass -File setup.ps1
Install cloudflared if needed: winget install --id Cloudflare.cloudflared
Manual (no setup script)
python3 local-daemon.py # auto-generates token if none given
cloudflared tunnel --url http://127.0.0.1:8765
Then paste URL + token to the agent.
Full protocol and security model: references/local-component-spec.md.
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 · 135 lines · 48 tokens per session scan A a191090dd088
desktop-subagent-connector is a skill published in the GitHub repository Stijnman/grok-custom-skills (9 stars, last pushed yesterday), licensed MIT. It adds 48 tokens to every session and 1,362 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
agent-browser
Use the host-side agent-browser CLI for local browser smoke tests, screenshots, snapshots, and simple UI validation against forwarded localhost URLs.
wechat-article-video-download
Download a specified video from a public WeChat Official Account article and verify the saved MP4. Use when the user provides an mp.weixin.qq.com article URL and asks to save, extract, grab, or download the first or Nth embedded article video. Do not use for Channels livestreams, login-only media, DRM, encrypted…
control-ui
Drive and inspect a local web, IDE, or Electron UI with browser or CDP automation and evidence. Use for UI regressions, screenshots, accessibility snapshots, visual diffs, focus, keyboard, scrolling, resize, console, network, performance, or heap checks. Do not use for backend-only changes or for a screenshot without…
takomi-flow
Use when Codex needs to operate Google Flow as a reusable agent tool through safe browser automation: bootstrap/login a persistent Flow profile, smoke-check access without spending credits, prepare image or video generation request files, run guarded Flow generations, download assets, inspect results, or integrate…
high-fidelity-extraction
Advanced protocol for extracting granular social media and web intelligence using browser automation and DOM analysis.
goal-meta-skill
A guide for turning an unclear or complex task into a reviewable /goal instruction for coding agents.