Download for MacApple silicon, macOS 14+
Skill

firetiger-create-agent

@firetiger-oss Updated 2026-07-08

Use when creating a Firetiger monitoring agent, installing a prebuilt catalog agent, or configuring an agent's triggers and schedules — monitoring something automatically, scheduling recurring or one-off analysis, or wiring event-driven automation. Always use this skill to create or automate an agent — it carries the critical gotchas (prefer create_agent_with_goal and answer the planner, agents must be AGENT_STATE_ON to fire, triggers are separate resources with six config types, scheduled-agent-runs are one-shot not cron).

Install

With your agent

Install OMGSkills skill "firetiger-oss/claude-plugin:skills/firetiger-create-agent" into this agent's normal skills directory: call the OMGSkills MCP get_install_plan tool for your agent and run the script it returns (it installs only the pinned commit and verifies SKILL.md), and stop if MCP or the pin is unavailable. Do not overwrite or change existing skills without asking me.

Manual command

git clone https://github.com/firetiger-oss/claude-plugin /tmp/claude-plugin && ln -s /tmp/claude-plugin/skills/firetiger-create-agent ~/.claude/skills/firetiger-create-agent

From README

Firetiger Create Agent Firetiger agents autonomously analyze telemetry, detect anomalies, investigate issues, and take actions through connections (Slack, GitHub, Linear, databases, …). Three ways to get one — prefer a goal or a catalog template over hand-building. The agent-planner (a system agent, agents/agent-planner) configures the prompt, connections, and triggers from the goal, then returns the planner conversation. If the planner asked a question, you must answer it or the agent won't finish — reply with sendagentmessage on the plan session. Path A (recommended): create from a goal Call createagentwithgoal with a clear goal (optional title, agentid, waitforplan default true). Handle planner questions. The planner asks sparingly ("act first, refine later"), via a requestuserinput. Answer on the user's behalf from the codebase/context — or ask the user if genuinely unknown — by writing to the plan session: Confirm. When the planner completes, the agent flips to AGENTSTATEON.