The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local commands including git (to find changed files), gh (to view PR details), and npm (to install dependencies). It also executes the agent-browser CLI to perform automation tasks.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the agent-browser package globally using npm install -g agent-browser if it is not already present on the system.
[DATA_EXPOSURE_&_EXFILTRATION]: The skill accesses local configuration files such as .env, package.json, and AGENTS.md to extract the development server's port number. This data is used strictly for local navigation (localhost) and is not transmitted externally.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository, including filenames from git diff and content from AGENTS.md or package.json for port detection.
Ingestion points: Files like AGENTS.md, package.json, and .env are read using grep and cat. PR file lists are retrieved via gh pr view.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present when processing these files.
Capability inventory: The skill can execute shell commands via npm, git, gh, and agent-browser.
Sanitization: Port detection uses regular expressions to extract digits, providing a basic level of sanitization against injection in that specific field.

test-browser