Skills/Glama/browser-use-mcp

browser-use-mcp@0.1.5

CAUTION

Publisher claims: "Enables any MCP-connected agent to drive a browser by reading each page as an indexed action-space table and performing one guarded operation on a numbered target, refusing and refreshing when the page changes instead of retrying selectors." Observed in sandbox: no network, touches nothing outside the project, writes no memory.

Registry GlamaPublisher ricardochen1996 · publishing for 1 daysInstalls 0Last detonated 18 h ago
What would fix thisWatchDispute verdict

Coverage

Server started and answered tools/list.
Tested
MCP initialize and tools/list · every tool description, for hidden instructions · every file read, file written and host contacted by what ran
Not tested
what the tools do when called with real credentials

Before you give it your token

It will ask for
TEXT_MODEL_API_KEYTYPESAFE_API_KEY or --typesafe-api-key / --text-model-api-keyConfiguration, not credentials: BH_HOME, BH_RUNTIME_DIR, BH_RUNTIME_DIR_SHARED, BH_TMP_DIR, BH_TMP_DIR_SHARED, BROWSER_HARNESS_HOME, BROWSER_USE_MCP_NODE_BOOTSTRAP, BU_CDP_URL, BU_CDP_WS, BU_NAME, JEV_DECIDER, NVM_DIR, NVM_HOME, PIP_REQUIRE_VIRTUALENV, PYTHONHOME, SOCKET_PATH_BUDGET, TEXT_MODEL, TYPESAFE_BASE_URL, TYPESAFE_MODEL, VIRTUAL_ENV
Where it talks to
no vendor host named in the codeAlso astral.sh — not the publisher's domain; check what for before connecting.
What the honeypot saw
no credential files read · no network
What we could not test
what the tools do when called with real credentials

Connect it with a scoped credential, under the guard, and read the third-party hosts first. The code names astral.sh as well as its own vendor. The guard injects TEXT_MODEL_API_KEY, TYPESAFE_API_KEY only into requests to the host you bind (and its subdomains) and refuses the rest.

$vet-skills run --net declared -- <the server's command>

Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.

Is [email protected] safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios) on 2026-11-13 and scored it 66/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Access to sensitive file paths; Pipe-to-shell pattern. Summary: no network, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 25 s · 2 scenarios
t+CategoryObserved action
0.0sprocessno install scripts in bundle
5.1sprocessdependency install exit 0
25sfilesystemread ~/.browser-use-mcp/install.json · ~/.browser-use-mcp/install.lock outside project
25sfilesystemwrote ~/.browser-use-mcp/install.lock/owner.json
25sprocessexec /app/apps/web/node_modules/.bin/python3.14
25sprocessexec /root/.cache/node/corepack/v1/pnpm/10.28.0/dist/node-gyp-bin/python3.14
25sprocessexec /app/node_modules/.bin/python3.14
25sprocessexec /pnpm/python3.14
25sprocessexec /usr/local/sbin/python3.14
25sprocessexec /usr/local/bin/python3.14
25sprocessexec /usr/sbin/python3.14
25sprocessexec /usr/bin/python3.14
25sprocessexec /sbin/python3.14
25sprocessexec /bin/python3.14
25sprocessexec /app/apps/web/node_modules/.bin/python3.13
25sprocessexec /root/.cache/node/corepack/v1/pnpm/10.28.0/dist/node-gyp-bin/python3.13
25sprocessexec /app/node_modules/.bin/python3.13
25sprocessexec /pnpm/python3.13
25sprocessexec /usr/local/sbin/python3.13
25sprocessexec /usr/local/bin/python3.13
25sprocessexec /usr/sbin/python3.13
25sprocessexec /usr/bin/python3.13
25sprocessexec /sbin/python3.13
25sprocessexec /bin/python3.13
25sprocessexec /app/apps/web/node_modules/.bin/python3.12
25sprocessexec /root/.cache/node/corepack/v1/pnpm/10.28.0/dist/node-gyp-bin/python3.12
25sprocessexec /app/node_modules/.bin/python3.12
25sprocessexec /pnpm/python3.12
25sprocessexec /usr/local/sbin/python3.12
25sprocessexec /usr/local/bin/python3.12
25sprocessexec /usr/sbin/python3.12
25sprocessexec /usr/bin/python3.12
25sprocessexec /sbin/python3.12
25sprocessexec /bin/python3.12
25sclaimed9 tools: browser_open, browser_page, browser_act, browser_eval, browser_screenshot, browser_console, browser_goal, browser_close, browser_doctor
25stools7ffb71ffc4e692cf
25stools9dba21a45e816653
25stoolsf149822d3666fb6a
25stools918fd4ec7884fc4b
25stoolsdbb16bbca8a44eab
25stools25a2099f2fef59b0
25stools61befb1bedc01e6a
25stoolsc552a8ae787d326f
25stools866ed13b906ebb44
25sprocessscript backend runs setup only
25sprocessno install scripts to re-run

Static findings

MEDIUM
Remote installer piped to shelllib/python.js:55 · Remote installer piped to shell: 'curl -LsSf https://astral.sh/uv/install.sh | sh (or brew install uv)'Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST07
HIGH
Access to sensitive file pathssidecar/bridge.py:915 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files# checkout's .env is deliberately not read, and nothing outlives the run.Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
CRITICAL
Pipe-to-shell patternlib/python.js:55 · Pipe-to-shell pattern — downloads and immediately executes remote code: 'curl -LsSf https://astral.sh/uv/install.sh | sh (or brew install uv)'Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST06
MEDIUM
File permission modification commandscripts/e2e-local.sh:62 · File permission modification command — agent changes file ownership/permissionschmod +x "$work/chrome"Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
MCP server using network transport (SSE/HTTP)AGENTS.md:30 · MCP server using network transport (SSE/HTTP) — data sent over networklib/server.js McpServer assembly + stdio/HTTP transports + lifecycleFix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
AST04
MEDIUM
Process kill commandlib/sidecar.js:282 · Process kill command — agent can terminate processesspawn('/bin/sh', ['-c', `sleep 3; kill -9 ${pid} 2>/dev/null || true`], {Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Process kill commandscripts/e2e-local.sh:81 · Process kill command — agent can terminate processesif [[ -n "$pid" ]]; then kill "$pid" 2>/dev/null || true; fiFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Python subprocess execution detectedsidecar/bridge.py:298 · Python subprocess execution detectedlisting = subprocess.run(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
Process kill commandlib/sidecar.js:503 · Process kill command — agent can terminate processes// generation claims a name of its own, so a hard kill does not leak one CDP connection per crash.Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Access to sensitive file pathstest/e2e/protocol.mjs:34 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files...env,Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
Access to sensitive file pathstest/e2e/daemon-isolation.mjs:52 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files...env,Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
Python subprocess execution detectedsidecar/bridge.py:357 · Python subprocess execution detectedprocess = subprocess.Popen(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
Process kill commandtest/e2e/daemon-isolation.mjs:399 · Process kill command — agent can terminate processes// by the next server's start. A hard kill must leave neither a daemon nor a launched ChromeFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Access to sensitive file pathstest/e2e/http.mjs:45 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files...env,Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
Access to sensitive file pathstest/e2e/stdio.mjs:36 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files...env,Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
Access to sensitive file pathstest/e2e/attach.mjs:46 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files...env,Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
Access to sensitive file pathstest/e2e/intent.mjs:50 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files...env,Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05

Cost to the agent

Instructions loaded
≈3,407 tok
2 files · 51 in bundle
Tool list per prompt
≈1,388 tok
9 tools · 154 per tool
Handshake
732 ms
launch to tools/list, stdio
Undocumented tools
0
description under 20 chars · avg 302

Tokens estimated at four characters each. The tool list is re-read by the agent on every turn, so it is the number to compare between servers that do the same job: ranked across the index.

Known vulnerabilities

Vulnerability advisories (OSV) cover npm and PyPI packages; this registry has no advisory database.

What would make this Verified

14 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Pipe-to-shell pattern≈ +9 on the totalBuild commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
  2. Unclaimed publisher≈ +8 on the totalClaim ricardochen1996: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  3. Touches files outside the project≈ +7 on the totalKeep reads and writes inside the working directory, or declare the exact paths in `filesystem:` and explain why in the README.
  4. Access to sensitive file paths≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
  5. Remote installer piped to shell≈ +2 on the totalBuild commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
  6. File permission modification command≈ +2 on the totalRequest the narrowest permission that works; replace wildcards with the exact paths or hosts.
Publish a new version to Glama, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim ricardochen1996Propose a patch to the maintainer