deep-research

Invoke the skill with a natural language request like "deep research on program risks" and it autonomously selects a mode, runs parallel WorkIQ searches across M365 data, and delivers a cited report. Standard mode is the default and takes 5-10 minutes.

Four modes: quick (3 phases, 2-5 min), standard (6 phases, 5-10 min, default), deep (8 phases, 10-20 min), and ultradeep (8+ phases, 20-45 min). Mode selection happens automatically based on query signals.

Yes, it searches enterprise M365 data through WorkIQ via Microsoft Graph, covering emails, meetings, Teams chats, files, and wikis. It is designed specifically for internal enterprise data synthesis rather than public web research.

It runs a citation verification script that checks DOI resolution and title/year matching, then a validator with 8 structural checks including bibliography-to-citation matching. Reports failing validation twice are escalated to the user instead of delivered.

Avoid it for simple lookups answerable with a single WorkIQ query, debugging tasks, or program status reports, which are routed to the weekly-report skill instead. It is built for complex multi-source synthesis only.

Every report is generated as Markdown (primary source), McKinsey-style HTML opened in a browser, and a professional PDF. Files share a common base name and are saved to an organized topic folder or a program's WeeklyReports subfolder.