The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external diagnostic data including application logs, distributed traces (Jaeger, Tempo, Datadog), and Kubernetes event metadata.
Ingestion points: Processes data from external logging and APM platforms as described in SKILL.md and references/common-failure-areas.md.
Boundary markers: The instructions do not define specific technical delimiters for data ingestion but provide clear behavioral instructions to cross-check AI-summarized logs against raw evidence.
Capability inventory: The skill is authorized to use execute_command, write_file, edit_file, and create_file for diagnostic and remediation purposes.
Sanitization: The skill includes an explicit "RCE Guard" (ASI05) that instructs the agent never to construct commands or fix scripts using external content without strict schema validation.
[COMMAND_EXECUTION]: The skill utilizes the execute_command tool to perform troubleshooting tasks. Security risks are addressed via explicit instructions in the "Security Guardrails" section of SKILL.md and references/failure-modes-and-security.md, which mandate that diagnostic commands must not be derived from untrusted external data.

troubleshoot-service