AI UNI · INTRANET · authenticated as guest · entitlement public_only · last regenerated at build
How AI Uni works.
A living corporate directory for the agents, tools, learnings and research that build the company. Read this end-to-end to understand the team — its shape, its standing disciplines, the research that grounds its decisions, and the gaps it's honest about.
Chapter 1 · where we started
A small set of logical roles directing the work
Three primary logical roles direct work — the User sets strategy and final quality, the PMO holds cross-session continuity and bookkeeping, and the PO authors plans and decides routing. CC orchestrators execute: commits, PRs, dispatches.
The team grew from there. Specialized direct-instances spin up when a lane needs deep expertise — Designer for visual direction, UX for heuristic evaluation, QA for 5-mode testing, Researcher for landscape briefs. Each agent works in its own terminal when the work is substantive enough to warrant Pattern A direct-instance.
Chapter 2 · what we've built
Six surfaces · 13 agents · 30 skills · 26 hooks · 17+ research artifacts
People first — then the tools they use, then the principles those tools encode, then the research that grounds it.
Every surface reads from a single source-of-truth in docs/agent-knowledge/intranet-content/. The directory regenerates from the corpus daily. Counts on filter pills and headers all derive from the same utility — no hardcoded numbers anywhere.
Section 3 · directory
People / Agents
13 agent cards. Click any agent to see resume-style bio with research portfolio, project contributions, learnings authored, skills used.
→ open directorySection 2 · org chart
Org chart + relationship graph
Hierarchy view of the agent team — primary logical roles · specialized direct- instances · Track C deferred classes. Bidirectional agent ↔ skill ↔ hook edges.
→ open org chartSection 5 · skills + hooks
Tools & Skills
Skills the team's agents use — workflow protocols, evaluation rubrics, build pipelines. Plus mechanical-enforcement hooks that keep discipline in place.
→ open librarySection 8 · accumulated discipline
Learnings
Per-agent principles, R-numbered cross-cutting governance, and the session- corrections cluster (S58 → S75) — the team's accumulated discipline log.
→ open librarySection 9 · grounded evidence
Research
Research artifact catalog — Meta-Researcher world-class brief, per-agent expertise corpora, landscape briefs, Memory Architecture brief, future-artifacts pipeline.
→ open libraryChapter 3 · where we're going
What we don't have yet — explicit, not hidden
A directory that hides its gaps overstates the team. 4 corpora are in flight, 3 classes are pre-amendment, 1 Track C classes are deferred, and 3 infrastructure items are queued — surfaced from gaps.json single source of truth.
Tier 2 · in flight
QA · Validator · Code-PM · PO
L-5 / L-6 / L-7 / L-8 expertise corpora authored in parallel worktrees; profiles auto-populate when corpora land per the refresh strategy.
Pre-amendment
Pre-amendment classes
Classes whose M-1 amendment hasn't fired yet render with placeholder badges until the next M-1 Meta-Researcher amendment lands their expertise floor.
Track C · deferred
Architect · Security · Portfolio-PM
NEW classes authorized for the next M-1 amendment expansion (S76 P0-5). Surfaced explicitly on the org chart so the team's planned shape is honest.
Activity feed · v2+ scope