The Agent Skills Directory

[COMMAND_EXECUTION]: Provides instructions and snippets for executing powerful debugging utilities including pdb, delve, lldb, and node --inspect. It also documents process-level inspection using shared namespaces in Docker and Kubernetes environments.
[EXTERNAL_DOWNLOADS]: Recommends the use of widely adopted third-party tools and container images for debugging, such as ipdb, pudb, delve, and the nicolaka/netshoot network diagnostic image.
[PRIVILEGE_ESCALATION]: Documents the use of kubectl debug node/ which provides administrative-level access to a cluster node's filesystem and process space for troubleshooting purposes.
[INDIRECT_PROMPT_INJECTION]: The workflows for production debugging rely on the ingestion of external data such as application logs and error reports which may contain untrusted content.
Ingestion points: Processes logs and telemetry data from Sentry or OpenTelemetry collectors during root cause analysis (references/production-debugging.md).
Boundary markers: Lacks explicit boundary delimiters for untrusted log content, though it favors structured JSON logging.
Capability inventory: The agent has access to system-level debugging tools capable of memory manipulation, process termination, and network traffic capture.
Sanitization: The instructions include best practices for sanitizing production data before reproduction and avoiding the logging of sensitive information like PII or credentials.

debugging-techniques