# Research Workflowware

Research Workflowware is the shelf of packages that make agent/human research disciplined, tagged, evidence-backed, and reusable.

This shelf supports customer packages by turning market signals, pain language, examples, risks, and evaluation cases into durable artifacts.

## Primary internal/pro package

### Tehuti Research Kit

- **Status:** internal/pro MAAT artifact
- **MAAT ID:** `9f4e1017-ae33-40d1-86ab-8883605d7113`
- **Type:** `research_kit`
- **Purpose:** best research stack for MAAT/Hermes agents.

The kit bundles:

- Evidence Bank;
- Art of Research method registry;
- KA2 methodology;
- evidence tagger skill;
- MAAT evidence tagging CLI.

## Research operating rule

```text
find → tag → verify → bank
```

## Method discipline

- `methods_used[]` should use Art of Research registry IDs.
- KA2 is one method, not the umbrella for every research run.
- Source channels are separate from methods.
- Evidence should be recoverable, tagged, and tied to artifacts.

## Public/private boundary

This public document intentionally does not expose private server paths, raw evidence stores, internal scripts, credentials, client data, or private MAAT runtime endpoints.

Private MAAT holds the operational paths and full research kit until explicitly exported.

## Workflowware library role

Research Workflowware feeds:

- customer pain-language research;
- competitor maps;
- buyer-type analysis;
- sample/eval datasets;
- risk-case datasets;
- offer-copy validation;
- package QA.

For Workflowware revenue packages, research artifacts should be created before public package ZIPs whenever the workflow depends on customer language, market demand, or risk classification.