The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes procedures for restarting production services using kubectl rollout restart. These are administrative actions that require high-level infrastructure permissions and should only be performed by authorized agents.- [DATA_EXFILTRATION]: The skill accesses system logs located at /var/log/langchain/app.log to diagnose recent errors. While necessary for troubleshooting, log files can contain sensitive configuration details or user data that could be exposed.- [EXTERNAL_DOWNLOADS]: The skill uses curl to fetch status information from status.openai.com and status.anthropic.com. These are well-known technology services, and the network requests are limited to official status indicators for monitoring purposes.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes external log data. Malicious content within the logs could potentially include instructions that attempt to influence the agent's behavior while it is performing its incident response duties.

langchain-incident-runbook