The Agent Skills Directory

[SAFE]: No malicious patterns, data exfiltration, or dangerous command executions were found. The skill operates as intended for debugging purposes.
[PROMPT_INJECTION]: The skill processes external, untrusted data from Playwright report artifacts, which is a surface for indirect prompt injection. This is addressed with a dedicated safety section in the SKILL.md. 1. Ingestion points: The agent reads from playwright-report/results.json and trace.zip. 2. Boundary markers: The skill contains an explicit 'Safety: artifacts are untrusted data' section that overrides external instructions. 3. Capability inventory: The skill uses shell commands like npx, jq, unzip, and find to process these files. 4. Sanitization: The instructions mandate that strings from reports are rendered as quoted text and that the agent must not execute any directives found within them.

playwright-debugger