Agent Stuff is a personal package of reusable skills, extensions, prompt commands, themes, and utilities for the Pi Coding Agent. Its author uses these tools across projects, and the catalogue entries are components from this package that support that workflow.
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 mitsuhiko/agent-stuff --skill web-browsergit clone --depth 1 https://github.com/mitsuhiko/agent-stuffWrote 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/mitsuhiko/agent-stuff/web-browser)<a href="https://agentmods.dev/skills/mitsuhiko/agent-stuff/web-browser"><img src="https://agentmods.dev/badge/skills/mitsuhiko/agent-stuff/web-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/mitsuhiko/agent-stuff/web-browser"><img src="https://agentmods.dev/badge/skills/mitsuhiko/agent-stuff/web-browser.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- Socket warn
- Snyk warn
- 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 YARA Match · line 50 YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).Fix: Remove the malware payload or compromised file entirely. Investigate how it entered the skill and audit all other artifacts for additional indicators of compromise.
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 | $0.00064 | $0.01324 |
| Opus 5.5 | $0.00026 | $0.00530 |
| Sonnet 5.5 | $0.00013 | $0.00265 |
| Haiku 4.5 | $0.00006 | $0.00132 |
Grade A, and why
web-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 1mo 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 — 164 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Web Browser Skill
Minimal CDP tools for collaborative site exploration.
Start Chrome (Prefer Headless)
./scripts/start.js --headless # Recommended: isolated reusable profile
./scripts/start.js --headless --profile # Headless with a copy of your profile
./scripts/start.js # Visible browser window when needed
./scripts/start.js --headless --reset-profile # Clear cached profile before launch
Starts Chrome with remote debugging (default port :9222). Agents should use --headless by default because it is less disruptive and supports navigation, evaluation, screenshots, emulation, and logging. User extensions are disabled for headless launches; a copied profile still provides its cookies and other browser state, but extension-based features are unavailable. Headed launches continue to load extensions normally. Use headed mode only when a person needs to see or interact with the browser, such as for pick.js, manual authentication, or debugging a headless-specific difference.
The start script only reuses a running browser when its profile and launch settings match. Stop the running skill browser before switching between headless and headed mode. Changing the debugging port does not isolate the profile: startup refuses to modify a profile already used on another port and preserves Chrome's profile locks.
Profile behavior:
- Default mode uses:
~/.cache/agent-web/browser/fresh-profile --profilemode uses:~/.cache/agent-web/browser/profile-copy- The skill does not attach to your live Chrome profile directly
- If
:9222is already used by an unknown instance, start will fail instead of reusing it
If Chrome is installed in a non-standard location, set:
BROWSER_BIN=/path/to/chrome ./scripts/start.js --headless
Optional debug endpoint override:
BROWSER_DEBUG_PORT=9333 ./scripts/start.js --headless
Stop Chrome
./scripts/stop.js # Stop skill Chrome on :9222
BROWSER_DEBUG_PORT=9333 ./scripts/stop.js # Stop skill Chrome on :9333
./scripts/stop.js --all # Stop all skill Chrome instances
What ships with it
18 files 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.
- scripts/browser-processes.js 1.2 KB runs code
- scripts/browser-processes.test.js 1.0 KB runs code
- scripts/cdp.js 7.5 KB runs code
- scripts/devices.js 3.4 KB runs code
- scripts/dismiss-cookies.js 14 KB runs code
- scripts/emulate.js 3.5 KB runs code
- scripts/emulation-state.js 1.9 KB runs code
- scripts/eval.js 2.2 KB runs code
- scripts/logs-tail.js 1.9 KB runs code
- scripts/nav.js 1.9 KB runs code
- scripts/net-summary.js 2.5 KB runs code
- scripts/package-lock.json 756 B
- scripts/package.json 68 B
- scripts/pick.js 5.6 KB runs code
- scripts/screenshot.js 4.4 KB runs code
- scripts/start.js 7.7 KB runs code
- scripts/stop.js 1.9 KB runs code
- scripts/watch.js 7.2 KB runs code
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.
- 1mo ago Changed · +17 lines 10055dcc9ea9
- 1mo ago First seen · 147 lines · 64 tokens per session scan A 2647ee685e09
web-browser is a skill published in the GitHub repository mitsuhiko/agent-stuff (3,192 stars, last pushed 21d ago), licensed Apache-2.0. It adds 64 tokens to every session and 1,324 once invoked, about $0.0003 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 2026-08-30.
Other skills, from other repositories
debug-optimize-lcp
Guides debugging and optimizing Largest Contentful Paint (LCP) using Chrome DevTools MCP tools. Use this skill whenever the user asks about LCP performance, slow page loads, Core Web Vitals optimization, or wants to understand why their page's main content takes too long to appear. Also use when the user mentions…
interactive-login
Log a CLI in with a browser/device-code flow (aws sso, gh, glab, gcloud, …) and save the result to the keychain so later commands can request it through execute.credentials.
pinchtab-mcp
Use this skill when a task requires browser automation through PinchTab's MCP server connected to a remote browser instance. Covers navigation, element interaction, data extraction, form filling, multi-step flows, and session management via MCP tools.
azure-messaging-webpubsub-java
Build real-time web applications with Azure Web PubSub SDK for Java. Use when implementing WebSocket-based messaging, live updates, chat applications, or server-to-client push notifications.
cua-driver
Use Cua Driver for desktop or browser tasks that are awkward or unavailable through Bash/APIs, or when the user explicitly wants GUI interaction: app testing, visual bug reproduction, form filling, calendar entry, screenshots, and demo recording. Also covers Cua setup; not OpenAI Codex Computer Use or web research.
open-design-browser
Inspect and exercise local HTML design artifacts with the existing agent-browser CLI, including screenshots and keyboard or interaction checks. Use when rendered evidence helps author or review a design; excludes browser installation, remote services, and unrelated browser automation.