test-playwright
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents an indirect prompt injection surface as it ingests untrusted data from live web applications and has significant capabilities.
- Ingestion points: The skill reads untrusted content via
snapshot(DOM content),consolelogs, andrequests(network data) from the browser. - Boundary markers: The skill references an external
protocol-browser-anti-stalldocument for safety rules, but no explicit input sanitization or boundary markers are present in the provided instructions to prevent obedience to instructions embedded in web pages. - Capability inventory: The agent can write to the local file system (Phase 4: 'Fix the root cause'), execute arbitrary SQL commands via the Supabase MCP (
execute_sql), and run arbitrary JavaScript in the browser context (run-code). - Sanitization: No explicit sanitization or filtering of external content before processing is mentioned.
- [COMMAND_EXECUTION]: The skill uses shell commands to interact with the environment, including
gitfor diffing changes,grepfor finding component consumers, and the Playwright CLI for browser automation. - [DYNAMIC_EXECUTION]: The skill utilizes Playwright's
evalandrun-codecommands to execute JavaScript directly within the browser context during the testing phase. - [EXTERNAL_DOWNLOADS]: The skill downloads and executes the Playwright CLI at runtime using
npx --yes @playwright/cli@latest. This fetches code from the official npm registry (a well-known service).
Audit Metadata