browserstack
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Executes shell commands to run tests (
npx playwright test) and install local testing agents (npm install). These are standard operations for a testing skill and target official packages from a well-known testing service.\n- [EXTERNAL_DOWNLOADS]: Downloads thebrowserstack-localpackage from the official npm registry, which is an expected dependency for the stated functionality.\n- [INDIRECT_PROMPT_INJECTION]: The skill reads external logs and session metadata from BrowserStack builds. This presents a theoretical surface for indirect prompt injection from test output, though it is a necessary function for testing tools.\n - Ingestion points: BrowserStack MCP tools (
browserstack_get_logs,browserstack_get_session) used inSKILL.md.\n - Boundary markers: None explicitly defined in the prompt instructions.\n
- Capability inventory: File modification (
playwright.config.ts) and shell command execution (npx playwright test) across the skill.\n - Sanitization: Not specified for ingested logs.\n- [SAFE]: The skill manages sensitive credentials (
BROWSERSTACK_ACCESS_KEY) by directing the user to set environment variables and providing the official link for credential retrieval, which is a secure implementation pattern.
Audit Metadata