The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs several shell operations including project building (pnpm build), starting a development server (pnpm dev), and database schema verification (npx prisma). While standard for testing, these represent high-privilege actions.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install external libraries (@browserbasehq/stagehand, zod) and toolsets (playwright) from the NPM registry.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from project documentation files to determine test logic and drive browser interactions.
Ingestion points: Reads from PRD-*.md, USER-JOURNEY.md, and EPIC-*.md to define testing scenarios.
Boundary markers: Absent. There are no delimiters or instructions to ignore embedded commands within the ingested files.
Capability inventory: Access to shell commands (pnpm, node, npx), file system access, and AI-driven browser control via Stagehand.
Sanitization: Absent. Data from these files is directly interpreted by the model to generate and execute test steps.

tester