Borrowing it
Nothing to install: this file belongs to dip497/hivemind. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dip497/hivemind/main/.claude/skills/hive-browser/SKILL.mdgit clone --depth 1 https://github.com/dip497/hivemindWrote 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/dip497/hivemind/hive-browser)<a href="https://agentmods.dev/skills/dip497/hivemind/hive-browser"><img src="https://agentmods.dev/badge/skills/dip497/hivemind/hive-browser/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/dip497/hivemind/hive-browser"><img src="https://agentmods.dev/badge/skills/dip497/hivemind/hive-browser.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 3 findings, up to medium
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 →
- medium MCP Rug Pull · line 65 npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.Fix: Pin the version: npx @scope/[email protected]
- medium MCP Rug Pull · line 102 npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.Fix: Pin the version: npx @scope/[email protected]
- medium MCP Rug Pull · line 66 npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.Fix: Pin the version: npx @scope/[email protected]
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 | $0.00192 | $0.02106 |
| Opus 5.5 | $0.00077 | $0.00842 |
| Sonnet 5.5 | $0.00038 | $0.00421 |
| Haiku 4.5 | $0.00019 | $0.00211 |
Grade A, and why
hive-browser 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 — 181 lines — stays where its author put it; the contents beside it link to each section on GitHub.
hive-browser
You are an agent running inside a hivemind tile (a terminal on an infinite
canvas). The canvas can host Browser tiles — real Chromium web views the
user can see. This skill lets you drive one of those tiles: same pixels the
user is watching, controlled over the Chrome DevTools Protocol (CDP) using the
agent-browser CLI.
You are not spawning a hidden headless browser. You attach to the user's visible tile, act in it, and they watch it happen. That shared context is the whole point.
How it works (the mental model)
- Each Browser tile is an Electron
<webview>with its own webContents. - hivemind exposes a loopback CDP port (when enabled) and writes a discovery file listing the open browser tiles (tile id, frame, URL).
agent-browser --cdp <port>connects to that endpoint; you pick the tab that is the user's tile and drive it withopen/snapshot/click/fill/screenshot.
Step 0 — Preconditions (check these first)
Run these and reason about the output before doing anything else:
# Is the CDP bridge enabled, and where is the discovery file?
echo "targets=$HIVEMIND_BROWSER_TARGETS port=$HIVEMIND_BROWSER_CDP_PORT"
cat "$HIVEMIND_BROWSER_TARGETS" 2>/dev/null || echo "NO_DISCOVERY_FILE"
Interpret:
port=is empty /cdpEnabled:falsein the file → the CDP bridge is OFF. It is opt-in for safety (a debug port also exposes the app window). Tell the user to enable it: Settings (gear, top-right) → "Enable agent browser control" → Relaunch to apply (or setHIVEMIND_BROWSER_CDP=1in the environment before launch). Then retry. Do not try to work around this.NO_DISCOVERY_FILEortiles: []→ no Browser tile is open. Ask the user to open one (canvas hotkey7, or a frame's + → Browser), ideally in the same frame as your tile. Then re-read the file.tiles:has entries → good, continue.
Make sure the agent-browser CLI is reachable. Prefer a global install if one
exists; otherwise run it on demand with npx — it fetches and caches the native
binary on first use, so there's no global install and no extra permissions:
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 Changed · +5 lines 818bc6bceaf5
- 2mo ago First seen · 176 lines · 192 tokens per session scan A db3e90bf4565
hive-browser is a skill published in the GitHub repository dip497/hivemind (17 stars, last pushed yesterday), licensed MIT. It adds 192 tokens to every session and 2,106 once invoked, about $0.0008 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 2026-08-31.
Other skills, from other repositories
record-video
Record a short WebM video of a web page or web-based app with Playwright's page.screencast, demonstrating a feature or bug fix. Use when asked to record evidence, demo a change, show that it works, record a video, or capture a screencast.
Abu-Browser
Browser controls for Abu's built-in visible web browser. They let an agent open pages, click, fill in forms, take screenshots, and extract information.
browser-harness-authoring
Use when mapping a repeatable website workflow into a verified Hermes skill so later runs can follow known steps instead of rediscovering the site. Surveys browser compatibility, semantic targets, failure modes, recovery paths, decision gates, and expiry using dummy data and no irreversible submissions.
reddit-browse-and-post
Let an agent read Reddit by default, and publish only after the user explicitly opts in and approves the exact post. Everything here is account-agnostic: the user supplies credentials through environment variables or a browser login, never through chat.
browser-core-efficient
Use when working in the embedded browser and you want the cheapest reliable interaction loop.
browser-qa
Use when validating or debugging a workflow in the embedded browser and you need a reproducible, evidence-first loop.