Automation guides

How HeyRocky separates publishing, commenting, resharing, and AI News curation into small scheduled workers with safety rules.

How Hermes cron jobs let Rocky run recurring briefs, checks, and automations safely with self-contained prompts and delivery rules.

A practical guide to script-only cron jobs in a verified Hermes/Rocky workflow.

A practical guide to attach cron jobs to sessions in a verified Hermes/Rocky workflow.

A practical guide to delegation and subagents in a verified Hermes/Rocky workflow.

A practical guide to background processes in a verified Hermes/Rocky workflow.

A practical guide to long-running agent spawns in a verified Hermes/Rocky workflow.

How to create daily Rocky briefings that are worth reading by focusing on changes, urgency, approvals, and next actions.

How to design recurring Rocky jobs that alert only when useful, stay quiet when nothing changed, and avoid duplicate notifications.

How to design a daily Rocky brief that summarizes what changed, what is urgent, what needs approval, and what can wait.

A practical pre-flight checklist for turning a Rocky idea into a bounded automation with clear inputs, safe actions, rollback, and proof of completion.

How to design public-safe event-driven Rocky workflows with clear payload boundaries, test events, and delivery verification.

Design webhook-triggered Rocky workflows with authentication, payload limits, replay protection, and human-readable delivery reports.

A public-safe runbook for recurring Rocky jobs that gather facts, summarize changes, and report only when there is something useful.

Prepare webhook-triggered Rocky workflows with signing, idempotency, delivery routing, logging, and public/private boundaries.

Run scheduled Rocky work like a release process: inspect sources, edit only safe surfaces, verify locally, deploy only on green checks, and report receipts.

Choose and operate gateway, plugin, and shell hooks without turning lifecycle automation into an invisible or over-privileged failure point.

Design event-driven Rocky workflows with signed routes, narrow prompts, idempotency, rate limits, and least-privilege tool access.

Use Hermes no-agent cron mode for deterministic watchdogs, alerts, and metrics that should run cheaply and stay silent when healthy.

Use persistent goals for multi-turn missions, then tighten autonomous progress with verification criteria, scope boundaries, subgoals, and wait barriers.

Trace job storage, tick locking, fresh-session execution, provider triggers, delivery, recursion guards, and scale-to-zero scheduling.

Evaluate suggested automations as proposed cron jobs, preserve the right delivery origin, and accept only bounded work with explicit safety and verification rules.

Use blueprint templates to create repeatable cron jobs through guided or direct slot filling, then inspect the concrete job before its first scheduled run.

Let goal loops wait on tracked builds and processes instead of wasting turns, with durable barriers, completion notifications, and stale-wait recovery.

Use hermes send to deliver raw script, file, CI, and monitoring output through configured messaging adapters without an LLM.

Trigger sandboxed pull-request reviews from signed GitHub events, fetch the real diff, and post concise idempotent comments.

Keep a durable objective active across responses while capping judge-driven continuation turns and preserving pause, resume, and human steering.

Create reviewed no-model shortcuts for repeated shell actions across CLI and messaging surfaces, with narrow arguments, safe output, and normal approval boundaries.

Separate Kanban dispatch ownership from notification delivery so task updates return through the gateway profile that owns the originating adapter.

A practical pattern for parallel Hermes subagents that preserves context isolation, prevents file collisions, and requires parent-level verification.