The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill configures containers to mount the host Docker socket (/var/run/docker.sock), which provides full control over the Docker daemon and can be used to escape container boundaries.
Evidence: references/elk-stack-configuration.md and references/filebeat-configuration.md both specify the /var/run/docker.sock volume mount.
[DATA_EXFILTRATION]: The configuration explicitly targets sensitive system logs that may contain authentication attempts and security-relevant metadata.
Evidence: references/filebeat-configuration.md includes /var/log/auth.log in its log harvesting paths.
[COMMAND_EXECUTION]: The skill includes a deployment script that executes shell commands and orchestrates infrastructure via kubectl without input validation on the environment variable.
Evidence: references/log-aggregation-deployment-script.md executes kubectl apply and kubectl rollout based on user-provided arguments.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a pipeline for ingesting untrusted log data from various sources into the agent's context, creating an attack surface for indirect prompt injection.
Ingestion points: Logstash and Filebeat ingest data from /var/log/app/*.log, Docker container logs, and Kubernetes metadata.
Boundary markers: The Logstash configuration uses JSON and Grok filters to structure logs, but does not implement specific delimiters to prevent the agent from interpreting log content as instructions.
Capability inventory: The skill includes shell scripts (scripts/validate-config.sh), Kubernetes deployment capabilities (references/log-aggregation-deployment-script.md), and file writing templates.
Sanitization: Logstash filters are configured to drop debug logs, but no sanitization or escaping of log content is performed to prevent prompt injection.
[CREDENTIALS_UNSAFE]: Multiple configurations explicitly disable authentication and security features, exposing the logging infrastructure to unauthorized access.
Evidence: references/elk-stack-configuration.md sets xpack.security.enabled=false for Elasticsearch, and references/loki-configuration-kubernetes.md sets auth_enabled: false for Loki.

log-aggregation