The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references standard monitoring and logging libraries in its implementation examples.\n
Evidence: SKILL.md imports well-known Node.js packages including prom-client, pino, and @opentelemetry/api for metrics, logging, and tracing functionality.\n- [COMMAND_EXECUTION]: The skill provides incident management templates that include shell commands for system diagnostics and administration. These are standard tools for SRE workflows.\n
Evidence: SKILL.md includes example commands like kubectl get pods, kubectl rollout restart, and psql within the Runbook Template section.\n- [INDIRECT_PROMPT_INJECTION]: The runbook templates define workflows where an agent processes logs and command outputs. This creates a surface for indirect prompt injection if the output from these systems contains malicious instructions.\n
Ingestion points: Diagnostic steps in SKILL.md involving kubectl logs and database query outputs.\n
Boundary markers: Not defined in the provided markdown templates.\n
Capability inventory: The templates include service management capabilities such as scaling and restarting deployments in SKILL.md.\n
Sanitization: The templates do not specify methods for sanitizing or validating the data returned from system commands.

reliability-engineering