x-to-obsidian

Warn

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to install an external, untrusted package ('agent-browser') globally from NPM. The installation process includes an additional command 'agent-browser install' which may download further unverified binaries or scripts.
  • [COMMAND_EXECUTION]: The skill relies on executing shell commands using the 'agent-browser' CLI to perform sensitive operations such as opening URLs, capturing page snapshots, and executing arbitrary JavaScript within a browser context.
  • [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it retrieves and processes untrusted content from X/Twitter posts.
  • Ingestion points: Untrusted data is ingested via 'agent-browser snapshot' in 'SKILL.md'.
  • Boundary markers: There are no markers or instructions to isolate external content from the agent's logic, allowing potentially malicious text in a post to influence agent behavior.
  • Capability inventory: The skill has the capability to write markdown files to the user's local filesystem (Obsidian vault) and evaluate code in the browser.
  • Sanitization: There is no evidence of sanitization or validation of the extracted content before it is processed or stored.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 9, 2026, 10:07 PM