The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for administrative actions using kubectl, including modifying environment variables (kubectl set env), scaling replicas (kubectl scale), and managing secret objects (kubectl create secret). These are high-privilege operations required for the skill's incident response purpose.
[COMMAND_EXECUTION]: Includes multiple curl commands to interact with Apollo.io's official API and status endpoints, as well as local service endpoints for metrics and health checks.
[CREDENTIALS_UNSAFE]: References the $APOLLO_API_KEY environment variable and provides templates for rotating and updating API keys within Kubernetes secrets.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection. 1. Ingestion points: Data is ingested from curl responses (API status, metrics) and kubectl logs. 2. Boundary markers: No specific delimiters or instructions to ignore embedded commands are present when processing this data. 3. Capability inventory: The skill has extensive cluster modification capabilities via kubectl (SKILL.md). 4. Sanitization: Ingested data is filtered using jq and grep but is not sanitized to prevent instruction injection.

apollo-incident-runbook