Operatives-specific terminology. For shared A Player OS terminology, see the master glossary.
| Term | Usage |
|---|---|
| A Player Labs | The company. Always "A Player Labs", never abbreviated to "APL" in user-facing text. |
Three stages of A Player OS. Each stands alone as a name. Never prefix with "A Player".
| Term | Stage | What it does | Repo |
|---|---|---|---|
| Playbooks | Build | Builds software. Linear skill chain. | aplayerlabs/playbooks |
| Operatives | Run | The agent factory. Designs operatives, produces deployment-ready specs. This repo. | aplayerlabs/operatives |
| Debriefs | Learn | Compounds knowledge from what ran. | aplayerlabs/debriefs |
A Player OS: Build → Run → Learn → Build better. Assets sit at the center — the shared state all three stages act on.
| Term | Definition | Scope |
|---|---|---|
| Skill | A slash command. The unit of interaction. Each skill has a SKILL.md (metadata) and CLAUDE.md (operating contract). | All three products |
| Playfield | The project folder where skills read from and write to. The bounded environment. | All three products |
| Assets | The shared state at the center of A Player OS. Design tokens, operative specs, code, decision history, component libraries — everything that persists between cycles. No stage owns them. All stages shape them. | All three products |
| Operative | An autonomous specialist. Deployed into a playfield with orientation, values, and purpose. The thing you design using Operatives. | Operatives |
| Operation | A recurring compounding cycle. Defined by the Five Jobs: physical output, boundary event, signal, leading indicator, closing mechanism. | Operatives |
| Mission | An autotelic, goal-focused task that completes. Has objective and completion criteria. Findings feed operations. | Operatives |
| OPERATIVES.md | The deployment manifest. The spec. Lives in the playfield root. Each skill owns a section. When complete, contains everything a runtime needs to instantiate and run an operative. | Operatives |
| The Five Jobs | The method for defining an operation: (1) name the output, (2) find the boundary, (3) stress test the signal, (4) confirm it's leading, (5) close the cycle. | Operatives |
| Level | Term | Description |
|---|---|---|
| Atomic | Single operation | One recurring compounding cycle |
| Department | Operation network | Multiple connected operations sharing compounding |
| Organism | Operative network | The full system — multiple operatives across playfields |
| Slash Command | Role |
|---|---|
/operatives |
Front door router. Reads state, points the way. |
/significance |
Define your significant purpose. |
/throughline |
Check alignment with purpose. |
/operation-design |
Design a single operation (the Five Jobs). |
/signal |
Design instrumentation and decision memory. |
/orient |
Build 5-component agent orientation. |
/values |
Inject values into the system. |
/wins |
Evaluate which wins to compound. |
/update |
Pull the latest from GitHub. |
Every skill's CLAUDE.md starts with this pattern:
# CLAUDE.md — /[skill-name]
> **Authority**: This file is the operating contract for /[skill-name].
> It defines how this skill operates within Operatives.
These terms MUST NOT appear in any file (except this glossary and the lint script). The lint script checks for these on every push.
| Deprecated | Replacement | Notes |
|---|---|---|
| A Player Operatives | Operatives | Product name stands alone |
| A Player Loops | Operatives | Old product name |
| aplayerloops | operatives | Old repo/package name |
| aplayeroperatives | operatives | Intermediate mistake — never shipped |
| loop (as domain concept) | operation or mission | Operation if it recurs, mission if it completes |
| Loop Architecture | Operative Architecture | System-level design |
| loop-design | operation-design | Skill name |
| brain (as unit) | operative or skill | Operative = the agent, skill = the slash command |
| brains (as folder) | operatives | Folder name |
| LOOPS.md | OPERATIVES.md | State file |
Exception: "loop" as a programming construct (for loop, while loop) is fine. The lint script ignores these.