The Agent Skills Directory

EXTERNAL_DOWNLOADS (LOW): The skill requires the installation of the playwright Python package and the execution of playwright install chromium to download browser binaries. While these are standard tools for web testing, they involve fetching third-party code and executables at runtime.
PROMPT_INJECTION (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8) as it navigates to and processes content from external URLs provided at runtime.
Ingestion points: scripts/playwright_runner.py ingests untrusted data from the web via page.goto(url).
Boundary markers: Absent. There are no delimiters or instructions to ignore embedded commands within the processed page content.
Capability inventory: The skill allows Bash, Write, Edit, and Grep tools, which provide a significant attack surface if an agent is successfully subverted.
Sanitization: Absent. The script extracts page titles and content without sanitizing potential instructions.
DATA_EXFILTRATION (LOW): The script allows the agent to navigate to any URL. This creates a surface for Server-Side Request Forgery (SSRF), where an attacker could potentially use the agent to probe internal network services or cloud metadata endpoints (e.g., 169.254.169.254).

webapp-testing