test-exploratory

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting data from a live web application, which represents a potential attack surface for indirect instructions.
  • Ingestion points: The agent captures page snapshots, accessibility (a11y) trees, console logs, and network events from the target application (SKILL.md).
  • Boundary markers: There are no explicit instructions to the agent to treat application content as untrusted or to use delimiters when processing captured logs.
  • Capability inventory: The agent utilizes playwright-cli to perform navigation, clicks, and form submissions on the target site.
  • Sanitization: No sanitization or filtering of the captured application data is specified before the agent evaluates it for the final report.
  • [EXTERNAL_DOWNLOADS]: The skill uses npx to download and run the Playwright CLI at runtime.
  • Evidence: PW="npx --yes @playwright/cli@latest" in Phase 0 of SKILL.md. This fetches the package from the official NPM registry.
  • [COMMAND_EXECUTION]: The skill executes shell commands to drive browser automation and manage session states.
  • Evidence: Use of shell variables and command execution strings such as $PW -s=explore-guest open --headed http://localhost:3000 and profile management under ~/.playwright-cli-profiles/.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:27 PM