The Agent Skills Directory

[COMMAND_EXECUTION]: Multiple scripts within the skill and its reference files utilize subprocess.run to perform administrative actions. This includes Kubernetes operations via kubectl, service management via systemctl, and low-level network manipulation using iptables and tc for chaos engineering simulations.
[REMOTE_CODE_EXECUTION]: An automation script in SKILL.md implements a control loop that fetches metrics from a Prometheus API and uses the results to trigger a kubectl rollout restart. This pattern creates a dependency where external monitoring data directly influences high-privilege command execution.
[COMMAND_EXECUTION]: The AutomatedRunbook implementation in references/automation-toil.md uses subprocess.run with shell=True to execute runbook steps. This practice increases the risk of command injection if command strings are dynamically generated or modified by the agent or user input.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to reach a Prometheus instance (http://prometheus:9090). Although this is typically an internal service, the capability to fetch data that subsequently controls system state is a sensitive attack surface.
[PROMPT_INJECTION]: The skill processes external data (Prometheus metrics) to make automated decisions. This represents an indirect prompt injection surface where manipulated metrics could theoretically influence the agent's logic or the parameters passed to administrative scripts.

sre-engineer