| hub | workspace-hub — the orchestration repo; contains skills, work-queue, scripts, config |
| submodule | Any repo registered in .gitmodules under workspace-hub (24 total as of 2026-03); includes tier-1, site, data, personal repos |
| tier-1 repo | A submodule with a pyproject.toml, uv.lock, and full test suite; currently: assetutilities, digitalmodel, worldenergydata, assethold, OGManufacturing |
| adapter file | AGENTS.md, CODEX.md, GEMINI.md — thin per-agent configuration files (≤20 lines); never put logic here, use skills instead |
| harness | The tooling layer inside hub: skills, work-queue, scripts, config, rules |
| WRK | A work item tracked in .Codex/work-queue/ with a WRK-NNN identifier |
| stage | One numbered step in the 20-stage WRK lifecycle (1=Capture … 20=Archive) |
| phase | A grouping of related stages (e.g., "planning phase" = stages 2-7) |
| checkpoint | checkpoint.yaml snapshot written after each stage to enable safe resume |
| gate | A hard stop requiring human confirmation or verifiable script output before proceeding |
| cross-review | Multi-provider AI review (Codex + Codex + Gemini) — Codex is a hard gate |
| AC | Acceptance Criteria — the testable conditions that define WRK completion |
| evidence | Files in assets/WRK-NNN/evidence/ that record gate passage |
| skill | A SKILL.md file loaded on demand to guide agent behaviour for a specific task type |
| orchestrator | The AI agent that owns the session; delegates to subagents |
| subagent | Agent spawned by the orchestrator to isolate context or run parallel work |
