The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references standard, well-known Node.js packages for observability and testing, specifically from the OpenTelemetry project (@opentelemetry/sdk-node, @opentelemetry/exporter-trace-otlp-http, @opentelemetry/resources, @opentelemetry/semantic-conventions) and Microsoft (@playwright/test).
[INDIRECT_PROMPT_INJECTION]: The skill identifies production error logs and traces as sources for test design, creating a surface where untrusted data could enter the agent's context. While the provided examples are benign, standard security practices for handling external log data should be observed.
Ingestion points: The analyze-production-errors.ts script processes ProductionError objects which contain message and stack fields from external error trackers.
Boundary markers: No explicit boundary markers or 'ignore' instructions are provided in the illustrative code snippets.
Capability inventory: The skill does not demonstrate dangerous capabilities (like eval, exec, or unauthorized network requests) that could be triggered by the analyzed log data.
Sanitization: No specific sanitization or validation of the ingested error strings is demonstrated in the code examples.

observability-driven-testing