The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install packages from public registries, including open-computer-use via NPM and cli-anything-hub via PyPI.
[COMMAND_EXECUTION]: The skill requires the execution of multiple shell commands to install software globally (npm install -g), check versions, and modify system configuration files.
[PROMPT_INJECTION]: The skill creates a high-risk surface for indirect prompt injection. By enabling browser control (Playwright) and desktop automation (open-computer-use), the agent will ingest and process untrusted data from the internet and local applications.
Ingestion points: Browser navigation content and desktop screenshots/window data as described in the verification steps of SKILL.md.
Boundary markers: Absent. There are no instructions or delimiters provided to help the agent distinguish between trusted instructions and untrusted data from the browser or desktop.
Capability inventory: The agent gains the ability to write to the file system (opencode.json), install global system packages, and perform arbitrary interactions with the desktop and web browsers.
Sanitization: Absent. No filtering or validation mechanisms are mentioned for the external data ingested.

opencode-browser