The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the kernel browsers process exec and kernel browsers playwright execute commands to run diagnostic checks and scripts within the managed cloud browser environment. These operations are restricted to the browser session and are standard for debugging activities.
[DATA_EXFILTRATION]: Provides functionality to read system logs and configuration files (/var/log/supervisord.log, /etc/resolv.conf) from the cloud browser VM using kernel browsers fs read-file. This access is intended for localizing service failures and does not target sensitive user-host credentials.
[PROMPT_INJECTION]: As the skill ingests external data sources such as browser console logs, network telemetry, and page titles via the CLI, it possesses an attack surface for indirect prompt injection. A malicious webpage could attempt to influence the agent's behavior through these logs.
Ingestion points: Browser logs, telemetry events, and page titles ingested via kernel-cli (SKILL.md).
Boundary markers: None identified; logs are processed directly.
Capability inventory: File system reads, process execution in VM, and Playwright script execution.
Sanitization: None identified; the skill relies on standard CLI output.
[EXTERNAL_DOWNLOADS]: References official documentation and telemetry categories hosted on docs.onkernel.com, which is the vendor's own infrastructure.

debug-browser-session