The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to resolve and execute the vibium CLI binary from multiple potential paths, including ./clicker/bin/vibium and ./node_modules/.bin/vibium.
[DYNAMIC_EXECUTION]: The vibium eval "<js>" and vibium wait fn "<expression>" commands allow the agent to execute arbitrary JavaScript code within the browser context at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from the web while providing powerful interaction tools, creating a surface for indirect prompt injection.
Ingestion points: The skill reads untrusted external data via vibium go, vibium text, and vibium html (SKILL.md).
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are provided in the skill instructions.
Capability inventory: The skill possesses capabilities including JavaScript execution (vibium eval), UI interaction (vibium click, vibium fill), and file system writes (vibium storage -o, vibium record).
Sanitization: No evidence of input sanitization or validation of the ingested web content is present.
[DATA_EXFILTRATION]: The skill provides commands to export sensitive browser session data, including cookies and local storage, to local files (e.g., vibium storage -o state.json). It also supports capturing screenshots and recordings of browser activity.

browser