assay

Warn

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires installing the assay-ui package from PyPI and triggers a download of a Chromium browser build during its initial setup.
  • [COMMAND_EXECUTION]: The skill involves executing the assay command-line utility through a shell to process local web pages.
  • [DYNAMIC_EXECUTION]: The skill uses Playwright to open and interact with web pages in a Chromium browser, executing any JavaScript contained within those pages. The skill documentation explicitly notes that serving pages over loopback does not provide a sandbox for scripts, and interactions could trigger external side effects.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted local web pages and reports the output verbatim to the agent, which could allow instructions embedded in the web content to influence the agent's behavior.
  • Ingestion points: CLI output from the assay tool which reads local HTML and JavaScript files (SKILL.md).
  • Boundary markers: Specific instructions to report the tool's output verbatim as the last part of the reply to provide clear delimitation (SKILL.md).
  • Capability inventory: Shell command execution and potential file modification via follow-up agent actions (SKILL.md).
  • Sanitization: No evidence of sanitization, filtering, or escaping of the tool's output before it is reported to the LLM (SKILL.md).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 14, 2026, 04:34 PM