Agent Browser (MikeFaierberg)

AdvisoryAudited by Static analysis on May 6, 2026.

Overview

No suspicious patterns detected.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

The agent could act as the user on logged-in websites or save/use website credentials, which may affect accounts or private data if misused.

Why it was flagged

The skill explicitly instructs use of logged-in browser profiles and a password/auth vault. That is purpose-aligned for browser automation, but the artifacts do not bound which credentials or sites may be used, how passwords are protected, or when user approval is required.

Skill content
Logged-in sites → Tandem Browser `profile="user"` ... `agent-browser auth save <name> --url <url> --username <u> --password <p>` ... `agent-browser --profile Default` to reuse Chrome login state
Recommendation

Use a dedicated browser profile or test account, avoid saving passwords unless necessary, and require explicit user confirmation before login, submission, purchase, posting, or account-changing actions.

What this means

On authenticated or transactional pages, automated clicks and keypresses could submit forms, change settings, or trigger purchases if not supervised.

Why it was flagged

The documented commands can click, fill, press keys, and batch browser actions. This is central to the skill’s stated purpose, but users should notice that these actions can have real effects on websites.

Skill content
`browser act request={kind:"click"...}` ... `browser act request={kind:"fill"...}` ... `browser act request={kind:"press", key:"Enter"}` ... `agent-browser batch ["open url" "snapshot -i" "click @e1"]`
Recommendation

Keep browser actions user-directed, and pause for confirmation before final submit/confirm/pay/post/delete steps.

What this means

If a different or untrusted `agent-browser` executable is on the system path, the agent may run that local program.

Why it was flagged

The skill relies on an external local CLI executed through the environment, while the supplied artifacts include no install spec or bundled code to verify that binary’s provenance.

Skill content
`agent-browser` CLI (v0.26.0) — standalone, extra features ... Complex automation → `agent-browser` CLI via exec
Recommendation

Install the CLI only from a trusted source, verify the expected version, and avoid using this skill on systems where the `agent-browser` binary provenance is unclear.