Noreja's Agents and Skills
Noreja comes with ready-made agents and skills for Minerva. This page lists them: which agents you can choose in the chat, which ones work in the background, and which skills you can start yourself with /. How agents and skills work in general is explained in What Minerva Can Do; how to create your own in AI Hub: Agents, Skills, Prompts and Workspaces and Write Your Own Skill.
Permissions: choosing an agent in the chat needs the feature Chat, and the agent must be shared with you (Noreja's agents are shared with everyone). The AI Hub tabs Agents and Skills & prompts need Agent or Prompt with Read. Items shared with you also open their tab without the feature, for example Noreja's agents and skills, which are shared with everyone. See Create and Edit Roles.
Agents You Can Choose in the Chat
Below the input field, Agent shows who answers. You choose it when you start a chat; it can't be changed afterwards. See Chat with Minerva.
Agent | What it does |
|---|---|
Minerva | The general assistant every chat starts with. Reads your process data, company knowledge and the Noreja documentation, works with files in workspaces (Excel, documents, slide decks), searches the web with OpenAI and Anthropic models, highlights activities in the process graph, plans larger tasks and hands parts of them to sub-agents. It may use all skills, also new ones. |
Minerva (On-Premise) | A lean variant for the local model, which can take much less text at once. Reads process data, company knowledge and the Noreja documentation, with fewer steps per answer. No files, no web search, no sub-agents, no graph highlighting. It uses only three skills: finding the perspective you named, reading graph results and violation analysis. Choose a local model with it. |
Your company may add its own agents; they appear in the same list if they are shared with you.
Sub-Agents in the Background
Sub-agents can't be chosen in the chat. Minerva starts them for part of a larger task, and the steps of the answer show their name:
- graph-analyst: works through one self-contained analysis, for example comparing many variants or breaking down violations, and returns the findings.
- task-planner: turns a complex request into a plan with ordered steps before Minerva carries it out.
For a single task, Minerva can also create a helper sub-agent with a name of its own, for example variant-comparator. It only exists for that answer.
Skills You Can Start Yourself
Minerva loads a skill by itself when your question matches it; the answer then shows Loaded skill with its name. To use a skill for sure, type / at the start of your message and choose it. The list shows up to eight skills at a time; type the first letters of a skill to narrow it down.
With the agent Minerva, the / list shows all 13 skills from Noreja, plus your company's skills that you may see. They fall into three groups.
Analyse Your Process
Type | Skill | Use it for |
|---|---|---|
| Conformance violation analysis | Analysing Disarray, Negligence and Rework, leaving out allowed deviations, with their effect on time and cost. |
| Process KPIs, rates and shares | Correct case counts, durations, rework and batching figures, and shares with both the absolute numbers. |
| Analysing activity and relationship properties | How the values of a property are distributed, for example order values or suppliers. |
| Validating a graph finding against customer context | Checking a finding against your company knowledge in the Context Hub before calling something a bottleneck or a violation. |
| Build process-graph filters | Narrowing an analysis to some of the cases, for example a period, the most frequent variants or only cases with violations. |
Present the Results
Type | Skill | Use it for |
|---|---|---|
| Building slide decks | A PowerPoint deck with your findings, in your company's colours and logo. Minerva saves it in your workspace. |
| Mermaid diagrams | Flow charts, sequences, timelines or mind maps in the answer. |
Minerva's Own Groundwork
Minerva uses these skills to do its work correctly. They are in the list too, but you rarely need to start them yourself.
Type | Skill | What it covers |
|---|---|---|
| Resolving a perspective the user named | Finding the perspective or analyzer tab you named, and when a tab is analysed instead of its perspective. |
| Reading a process-graph result | Reading the figures of activities and relationships correctly, for example that unexpected ends usually mean cases still in progress. |
| Interactive process-graph highlighting | The underlined links in answers that light up activities and relationships in the process graph. |
| Planning and todo lists | The plan with open steps that Minerva shows for larger tasks. |
| Delegating to a configured sub-agent | Handing a self-contained part of the work to a sub-agent such as graph-analyst. |
| Creating an ad-hoc sub-agent | Creating a helper sub-agent for one answer. |
The short descriptions in the / list are written for the AI and are in English. With Minerva (On-Premise), the list shows only the three skills that agent uses.
The AI Hub tab Skills & prompts lists more Noreja items than the / list: besides the skills, it holds the system prompts of Noreja's agents and user prompts that Noreja uses internally, for example to give chats a title. These can't be started in the chat.
Change or Copy Them
- Noreja's agents and skills are maintained by Noreja and updated with new releases. They can't be changed or deleted in your company.
- Duplicate in the AI Hub creates your own copy of an agent, which you can change and share. Your company's skills with the same name can't replace a Noreja skill.
- To fix a format for your company, for example a monthly report, write your own skill: Write Your Own Skill.