The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to interact with cluster and system management tools.
Scripts in SKILL.md and references/automation-toil.md execute 'kubectl rollout restart' to remediate service errors.
Chaos engineering patterns in references/incident-chaos.md execute 'iptables' and 'tc' to simulate network failures and latency.
The AutomatedRunbook class in references/automation-toil.md uses shell=True to execute runbook steps.
[PRIVILEGE_ESCALATION]: Several scripts perform actions that typically require administrative permissions.
The SelfHealer class in references/automation-toil.md includes methods for restarting system services (systemctl restart) and deleting log files (find /var/log ... -delete).
Network manipulation commands in references/incident-chaos.md (e.g., 'tc qdisc add', 'iptables -A OUTPUT') require elevated network privileges.
[EXTERNAL_DOWNLOADS]: The skill interacts with external and internal endpoints to fetch metrics and documentation.
Fetches error rate data from an internal Prometheus server (http://prometheus:9090) to determine if auto-remediation is necessary.
Links to external documentation on the author's GitHub Pages site (jeffallan.github.io).
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface where external metrics influence automated infrastructure changes.
Ingestion points: Reads metrics data from Prometheus API responses in SKILL.md.
Capability inventory: Significant capabilities including kubectl, systemctl, iptables, and tc are available.
Boundary markers: The scripts use specific logic (e.g., float comparison) to limit the impact of external data.
Sanitization: Prometheus queries are URL-encoded.

sre-engineer