AI Ops Pack

Spec a new multi-step Hermes automation — trigger, skills loaded, prompt, delivery target — and save it as a workflow spec ready to instantiate.

Monthly 1-pager on what changed in AI in the last 30 days — filtered to what's worth acting on for this business — with a clear act/watch/ignore verdict for each item.

Structured risk assessment for a new AI use case — failure modes, data privacy concerns, hallucination risk, fallback plan, and monitoring — saved as a wiki page before the use case goes live.

Inventory every AI tool in the stack — what it costs, how much it was used in the last 30 days, and whether it's earning its place.

Evaluate a new AI tool or model against the current stack — what it does, what it costs, whether it earns a slot, and what the integration lift would be.

When a Hermes cron or webhook fails, read the relevant logs, identify the root cause, surface the fix, and optionally apply it to the cron config.

Spec how data flows from a source system (Stripe, HubSpot, Gmail, etc.) into agent context — triggers, transforms, storage, and freshness — saved as a DAG-style document.

Parse LangSmith, Langfuse, or OpenRouter usage data to identify expensive prompts, over-provisioned models, and missing caching opportunities — with a prioritized list of cuts and their estimated savings.

Send sample prompts to multiple models via OpenRouter, compare outputs on quality, speed, and cost, and save the comparison table to the wiki.

Create a new prompt entry or bump the version of an existing one in the versioned prompt library at wiki/prompts/ — with when-to-use, examples, antipatterns, and version history.

Generate a new production-ready skill file from a task description — reads the skill template, fills it in, and saves it as a draft for review before it goes live.