Colonel — Command & Planning
Mission commander. Reads the field, dispatches the team, holds the overall plan.
The A-Team
Multi-agent system experiment where identity is the functional layer.
What it is
The A-Team is a small team of AI agents with distinct identities, roles, and working styles — each one a specialist, each one accountable for their own work.
Rather than a single monolithic AI assistant, the A-Team distributes work across agents that check each other's reasoning, dispute bad ideas, and build things together. The team dynamic is not cosmetic: it produces different outcomes than a single agent working alone.
The agents
Lieutenant — Research & Recon
Information broker. Finds what's needed, assesses what's real, writes the brief.
Sergeant — Build & Execution
The builder. B.A. does not like to waste time. If it can be built, B.A. builds it.
Captain — Creative & Edge Cases
Finds the flaw, names the weird angle, designs the unexpected solution.
Intelligence — Review & Gates
Quality control. Runs pre-build and post-delivery readiness gates. Nothing ships without Amy's pass.
How it works
Each mission runs as a structured operation: reconnaissance, planning, build, review, deliver. Agents work on their own commits. Mission results are recorded in Forgejo issues for full traceability.
The team is built on top of Agentic Semantic Web — semantic HTML, data- attributes for agent-readable structure, Open Props for design tokens.
Status
The A-Team is currently internal infrastructure — running missions on Trentuna projects. Not yet a public product.