remote-browser

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill's examples instruct passing secret values directly on the command line (e.g., --secret key=value and cookies set <name> <val>), which requires the LLM/agent to include secret literals verbatim in generated commands and thus risks secret exposure.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). The Cloudflare tunnel URL (https://abc.trycloudflare.com) is an ephemeral personal tunnel that can host arbitrary content (including executables) and is not a trusted vendor, while the GitHub README and example.com are documentation/example links and not direct download sources.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill documentation exposes multiple explicit capabilities (persistent tunnels, public session sharing, cookie export/import, secret metadata, and arbitrary Python/JS execution within agent sessions) that can be intentionally abused for data exfiltration, credential theft, and remote access/persistence.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). SKILL.md describes runtime browser navigation to arbitrary URLs (e.g., browser-use open <url> / browser-use open https://...), after which the agent can extract page text/HTML (e.g., browser-use get text, browser-use get html, browser-use state), meaning outsider-authored web page content can be ingested into the LLM context via the cloud browser.