| name | mopp-doctrine |
| description | Explains the MOPP (Mission-Oriented Protective Posture) model for AI work — the 0-4 levels, the cumulative gear/control stack, the dev and security faces, decision authority, and why de-escalation matters. Use when the user asks what MOPP is, how the levels map to guardrails, or which controls apply at a given level. |
MOPP doctrine
MOPP scales AI guardrails to the assessed threat (blast radius + sensitivity) of
the work, exactly like military protective gear scales to a chemical threat.
Higher posture = more protection, but real velocity cost — so you de-escalate
when the risky work is done rather than holding maximum protection.
Read the canonical model from core/doctrine.md (resolve via $MOPP_BIN's repo,
or the monorepo core/ directory). Key points:
- Five levels, cumulative. Each level keeps every control below it and adds
one gear layer: 0 base → 1 overgarment → 2 overboots → 3 mask → 4 gloves.
- Two faces, one level. Each level has a development control set and a
security control set. Same posture number, both stacks active.
- The mask (level 3) is where a human enters the loop. Below 3 the agent
runs on static gates; at 3+ a person approves destructive/sensitive actions.
- Commander dictates. The agent recommends a level via
mopp assess; the
human adopts it via mopp set. Never silently change posture.
Control summary (cumulative):
| Level | Dev face | Security face |
|---|
| 0 | full autonomy, branch only | none |
| 1 | tests required, branch isolation | baseline secret scan |
| 2 | lint/typecheck gate, review | dependency audit, SAST |
| 3 | approval on destructive ops, no force-push | egress confirm, audit log ON |
| 4 | worktree, dry-run, step confirm | least privilege, secrets out of context, rollback plan |
To inspect levels at runtime: node $MOPP_BIN explain <0-4>.
