@agimon-ai/po-browse-tool@0.0.0-alpha.4
MCP server for browser automation using Playwright with profile management, page registry, and multi-browser support
AI Security Review
scanned 19d ago · by lpm-firewall-aiThe manifest has no install lifecycle hook, and the inspected runtime behavior is consistent with browser automation and browser setup. Static flags for process output and networking did not establish an unrelated data transfer or other attack.
Decision evidence
public snapshot- package.json lists development and build scripts, with no install lifecycle hook; its entrypoints are the CLI and runtime modules.
- Both runtime entrypoints restrict browser launch to the bound profile and browser type.
- The Chrome installer requests version metadata from Google's Chrome for Testing feed and uses a cache for the browser executable.
- The extension browser code captures child-process output and uses it in a browser startup failure diagnostic.
- The extension bridge sends selected browser tool calls and their arguments to its configured execution endpoint.
- The flagged command-output behavior is tied to browser startup diagnostics; the inspected code does not show it sending captured output to an unrelated recipient.
Source & flagged code
7 flaggedPackage source references child process execution.
dist/dockerChromeForTesting-BMmEQq88.mjsView on unpkg · L1A single source file combines environment access, network access, and code or shell execution; review context before blocking.
dist/dockerChromeForTesting-BMmEQq88.mjsView on unpkg · L1Findings
Affected versions and remediation
This report applies to @agimon-ai/po-browse-tool@0.0.0-alpha.4.
See all versions of @agimon-ai/po-browse-tool for other recorded verdicts.
- Review the evidence and your use of @agimon-ai/po-browse-tool@0.0.0-alpha.4 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.