pw-debug-live

Pass

Audited by Gen Agent Trust Hub on Jun 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill frequently utilizes the Bash tool to execute shell commands for running tests and managing browser sessions via npx.
  • [EXTERNAL_DOWNLOADS]: The skill invokes npx playwright-cli, which downloads and executes a third-party package from the public npm registry. This package is not maintained by the official Microsoft Playwright team.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted data from browser snapshots and console logs without adequate isolation. * Ingestion points: Data entering through snapshot, console, and requests commands in Step 7. * Boundary markers: Absent; there are no instructions to the agent to disregard embedded content as instructions. * Capability inventory: Access to shell execution (Steps 2, 3, 10) and file editing tools (Step 8). * Sanitization: None; browser output is processed directly to diagnose test failures.
  • [REMOTE_CODE_EXECUTION]: The run-code feature allows the agent to generate and execute arbitrary Playwright TypeScript code against the live browser session, creating a dynamic execution risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 29, 2026, 11:30 PM
Security Audit — agent-trust-hub — pw-debug-live