The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The package.json file contains a postinstall script that executes npx playwright install chromium. This command downloads the necessary browser binaries from Microsoft's official Playwright distribution servers to enable the skill's automation features.
[COMMAND_EXECUTION]: The skill performs automated form filling and submission using the Playwright framework. This behavior is consistent with its stated purpose of dynamic security testing and is explicitly documented in the user instructions.
[PROMPT_INJECTION]: The skill ingests untrusted data by reading the HTML content of target web pages (src/runner.ts) to analyze for payload reflections. While this represents a surface for indirect prompt injection, the core logic is implemented in TypeScript rather than being driven by prompt-based instructions.
Ingestion points: Page content is retrieved via page.content() in src/runner.ts to identify reflected XSS payloads or SQL error strings.
Boundary markers: There are no specific delimiters defined in the instructions to separate the ingested page content from the agent's instruction stream.
Capability inventory: The skill has the ability to write files (screenshots in src/runner.ts), perform browser automation, and initiate network requests to target URLs.
Sanitization: The skill does not sanitize page content, as its intended function is to detect the raw presence of its own security payloads in the page response.

playwright-security-runner