The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends running npx playwright test, which may download the Playwright package and its browser binaries from official Microsoft/Playwright registries.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to process test results, including jq for JSON parsing, unzip for extracting trace archives, and node -e for running inline JavaScript logic to filter and display test errors.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests error messages and console logs from results.json and trace.trace files for failure classification.
Ingestion points: playwright-report/results.json and trace.zip (specifically trace.trace and trace.network).
Boundary markers: Absent. The skill provides structure via markdown and tables but does not use explicit delimiters or warnings to ignore instructions within the ingested data.
Capability inventory: File reading (cat, find, unzip) and script execution (node, jq).
Sanitization: None. The skill extracts and prints raw error messages and console output directly to the agent for analysis.

e2e-test-debugger