Hermes AgentUse-Case Playbooks

Each guide is self-contained: what the use case is, which agent primitives make it possible (memory, skills, cron, webhooks), and exactly how to set it up.

• 01 — Profiles & Memory Budget: isolated agent personas and the bounded-memory config everything else relies on
• 02 — Cost & Review Monitoring: the wakeAgent gate — a cheap pre-run script that only wakes the LLM when something actually changed
• 03 — System Health Checks: LLM-free cron watchdogs for TLS, uptime, and services — effectively free
• 04 — Second Brain over Slack: one shared, ever-learning source of company context for the whole team
• 05 — Auto-Respond to Leads over Gmail: lead detection, drafted replies, and meeting booking — including the full Google OAuth walkthrough
• 06 — Competitor Watchdog: a weekly cron that studies competitors and keeps your PRD current, using the PRD-as-a-skill pattern
• 07 — Social Media Drafting & Posting: video script → X + LinkedIn drafts → human review → publish via the xurl CLI

Most scheduled AI automations burn tokens on every tick whether anything happened or not. These playbooks are built around two mechanisms that fix that: the wakeAgent gate (a $0 pre-check that decides if the model should fire) and no-agent mode (the script's output IS the message — no model call at all). If you run agents on a schedule, this pattern alone pays for the read.

Anyone running Hermes Agent (by Nous Research) or designing scheduled agent automations on any platform — the gating patterns, bounded-memory philosophy, and skills-as-procedural-memory approach transfer directly.