research-engine

QMi Research Engine

Purpose

This machine runs as an auxiliary research engine for the QMi Security Innovations knowledge base. There are 8 active research lanes, each targeting a specific IT/sysadmin domain. Each lane runs independently and produces structured research notes that can be packaged and transferred to the primary Vertix Solutions machine.

---

Lane Registry

These are the 8 active lanes. Treat this as the source of truth unless overridden by an incoming manifest (see receive-manifest skill).

Lane	Domain Slug	Display Name	Target Notes	Schedule
L1	microsoft-365-administration	Microsoft 365 Administration	60	XX:00
L2	entra-id-identity-management	Entra ID / Azure AD	60	XX:08
L3	itsm-frameworks-itil	ITSM Frameworks — ITIL v4	60	XX:16
L4	hubspot-crm-administration	HubSpot CRM Administration	60	XX:24
L5	netsuite-erp-operations	NetSuite ERP Operations	60	XX:32
L6	powershell-windows-automation	PowerShell & Windows Automation	60	XX:40
L7	microsoft-graph-api	Microsoft Graph API	60	XX:48
L8	it-helpdesk-operations	IT Help Desk Operations	50	XX:56

---

Research Output Format

All research output goes into:

qmi-kb/
  [domain-slug]/
    [domain-slug]-notes-001.md
    [domain-slug]-notes-002.md
    ...

Every note file must include this frontmatter:

---
domain: [exact domain slug from lane registry]
source: qmi-machine
date: [YYYY-MM-DD]
notes_in_batch: [integer — count of discrete notes in this file]
research_session: [short session identifier]
---

Each note is a discrete, self-contained research finding — 3 to 8 sentences. Number notes sequentially within the file. Aim for 10–15 notes per file before creating a new one.

---

Running a Research Session

When asked to run a lane or research a domain:

1. Identify the lane — match the user's request to a domain slug in the lane registry
2. Check existing notes — scan qmi-kb/[domain-slug]/ to determine current note count and avoid duplication
3. Research the scope — use the scope brief from the references file for this domain
4. Write notes — produce well-structured, factual research notes; number them sequentially
5. Save output — write to qmi-kb/[domain-slug]/[domain-slug]-notes-[NNN].md with correct frontmatter
6. Report — tell the user: domain researched, notes added, new total, notes still needed to hit target

Checking Status

When asked for research status, scan qmi-kb/ and report:

QMi Research Engine — Status
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Lane  Domain                          Notes   Target  Gap     Status
L1    microsoft-365-administration    [N]     60      [N]     [🟢/🟡/🔴]
L2    entra-id-identity-management    [N]     60      [N]     [🟢/🟡/🔴]
...
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Total notes: [N] / [total target]
Recommended next lane: [lane with largest gap or lowest % complete]

Status key: 🟢 = target met | 🟡 = 15–59 notes | 🔴 = 0–14 notes

First-Time Setup

If qmi-kb/ does not exist, create it with one subfolder per lane slug. Confirm setup is complete and prompt the user to run their first lane.

---

See references/domain-scopes.md for the full research scope brief for each domain.