The Great Library of SISO · Agents · System map
One shared task brain. Clear homes for everything around it.
A direct source comparison of OS Database, Task Manager, PM Tasks, and Agent Brain. This dated projection can evolve; stable Works and exact release receipts preserve what was believed and why.
Agent Brain owns shared task truth. Project OS owns planning structure. Runtime owns session execution. Skills Hub owns capability discovery and thin adapters. Playbooks own composed procedures. The old mixed database folder is decomposed rather than promoted wholesale.
Where the source goes
| Legacy behavior | Canonical home | State |
|---|---|---|
| Task lifecycle, role queue, atomic claim, versioned task artifacts | SISO Agent Brain 0.3.0 | Implemented and tested |
| Task Manager commands | Skills Hub thin adapter | No database, no raw SQL |
| PM Tasks | Temporary Skills Hub alias | Retire after consumer migration |
| Projects, missions, goals | Project OS | Reclassified |
| Agent sessions and tool execution permissions | Agent Runtime | Reclassified |
| Skill and tool catalog | Skills Hub | Reclassified |
| Templates and multi-step pipelines | Agent Playbooks | Reclassified |
| Scheduled automations | Future Agent Automation Work | Deferred |
| Custom fields and generic relationship graph | Unverified | No promotion without consumer evidence |
| Raw SQL and machine-local database paths | None | Retired |
Receipts, not vibes
Agent Brain’s verifier creates isolated databases, applies three immutable migrations twice, races two workers for one step, versions artifacts, replays offline writes, and runs dry-run/apply/idempotent imports for both reviewed legacy layouts. Private path artifacts never copy; memory and timeline require explicit opt-in. Skills Hub proves its adapters route stable Brain commands, its telemetry writes only to a Skills-owned store, and Task Commander contains no raw SQL or topology snapshot.
Read the machine assessment, full human decision, consumer inventory, or Skills repository map.
Not yet claimed: dry-run review of a real local legacy database, consumers outside the published Skills Hub, sustained multi-host concurrency, or the value of unconsumed custom-field and graph features.