AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Agent Browser

skill-ahmadrosid-nakama-agent-browser · by ahmadrosid

Use for login-walled portals and headless browser automation with agent-browser. Skip public HTTP reading and explain-only questions.

— No reviews yet
0 installs
0 views
— view→install

Install

$ agentstack add skill-ahmadrosid-nakama-agent-browser

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • ✓ Prompt-injection patterns
  • ✓ Secret / credential exfiltration
  • ✓ Dangerous shell & filesystem operations
  • ✓ Untrusted network calls
  • ✓ Known-malicious package signatures

What it can access

  • ✓ Network access No
  • ✓ Filesystem access No
  • ✓ Shell / process execution No
  • ✓ Environment & secrets No
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-ahmadrosid-nakama-agent-browser)

Reliability & compatibility

✓ Security review passed
0 installs to date
— no reviews yet
● today

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Agent Browser? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Use this skill when the task needs a real browser: login walls, forms, multi-step UI, or pages that web_fetch / web_search cannot reach.

Keep ordinary web work local:

  • Use web_fetch for a single public HTTP(S) page that does not require login or interaction.
  • Use web_search to discover sources.
  • Do not open a browser just to read a public doc page.

Prerequisites

  • This profile must have the bash tool assigned.
  • The host must have the agent-browser CLI and Chrome installed:
npm install -g agent-browser
agent-browser install

If bash returns command not found, ENOENT, or similar for agent-browser, tell the operator to run the install commands above (and agent-browser install --with-deps on Linux if Chrome libraries are missing). Do not invent a different browser tool.

Optional: CloakBrowser (stealth Chromium)

Stock Chrome from agent-browser install is the default. Use CloakBrowser only when the site blocks that Chrome (bot detection / antibot). Cloak is not the default and is not bundled with Nakama.

The operator installs Cloak on the same host that runs bash and points agent-browser at Cloak's binary:

# Host env for the Nakama process (inherited by bash)
export AGENT_BROWSER_EXECUTABLE_PATH="/path/to/cloak-chromium"
export AGENT_BROWSER_ARGS=""

Get the binary path with python -m cloakbrowser info or npx cloakbrowser info after cloakbrowser install. Copy AGENT_BROWSER_ARGS from Cloak's agent-browser integration. Do not invent flags.

If those variables are unset, keep using stock Chrome. Do not pass Cloak paths in each bash env unless the operator asked you to override the host env for this run.

Browser workflow

Drive the browser with the bash tool. Multiple agent-browser commands in the same agent run share one daemon session — that is how login → navigate → act stays coherent.

  1. Open the target (or launch without a URL if you must set cookies first):
agent-browser open https://example.com/login
  1. Snapshot for token-efficient element refs (prefer interactive/ref output):
agent-browser snapshot -i
  1. Act by ref from the snapshot (@e1, @e2, …). Re-snapshot after navigation or major UI changes — stale refs fail:
agent-browser fill @e3 "user@example.com"
agent-browser fill @e4 "secret"
agent-browser click @e5
agent-browser press Enter
  1. Wait when the UI is loading:
agent-browser wait --load networkidle
  1. Optional screenshot — write under the profile workspace artifacts/ directory (bash cwd is the profile workspace). Prefer meaningful names; avoid capturing password fields:
agent-browser screenshot artifacts/browser-page.png
  1. When the task is done (success or give-up), close the session so the next run starts fresh:
agent-browser close

If the daemon looks stuck, agent-browser close --all or agent-browser doctor can help.

Timeouts

Default bash timeout is short. For browse/login steps, pass an explicit timeoutMs (tens of seconds to a few minutes per command as needed). Prefer one focused command per bash call so failures are easy to read.

Session policy (v1)

  • Within one run: reuse the same agent-browser daemon (do not close between every click).
  • Across runs: do not use --restore, state load, or persistent Chrome profiles unless the user explicitly asks in this conversation. Re-authenticate when the task requires it.
  • Always close at the end of the browser work in this run.

Credentials

  • Prefer credentials the user supplied in the prompt or already stored for this task.
  • Do not echo passwords or session tokens in your final summary.
  • Avoid screenshots of password fields.

After the run

  • Summarize what you did and what you learned in plain language.
  • If a command failed (non-zero exit, timeout, missing CLI), explain clearly and either retry with a fresh snapshot or ask the user / operator for help.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.