Logic & Planning Agent
Structured reasoning and plan decomposition.
- Doctrinal basis
- Supports mission analysis within NATO OPP / AJP-5.
- Deployment
- Runs alongside the planning agents in the same enclave.
- Underlying products
- SynapseMesh™VerityGridAI™
Purpose
Staff products can look coherent while containing contradictory tasks, missing dependencies or constraints that silently invalidate the plan. SC-LNP-01 decomposes commander intent into tasks, dependencies and constraints, then checks candidate plans for logical consistency before staff time is spent on them. The core parses a formal task graph, locates contradictions and gaps as specific, located defects, and continuously validates intermediate products from the planning agents. It does not generate plans of its own; it is the sceptic in the mesh, ensuring that what is built is internally consistent and traceable to intent.
Doctrinal basis
Supports mission analysis within NATO OPP / AJP-5.
Architecture
Commander intent, task organisation and the stated constraints, restraints and assumptions are parsed into a formal task graph. The core is a structured reasoner, not a free-text generator: it builds decomposition trees, checks dependencies and sequencing, and searches candidate plans for contradictions and gaps before staff time is spent on them. Findings are reported as specific, located defects, naming the tasks and constraints in conflict, rather than as general commentary. It runs in the same enclave as the planning agents and validates their intermediate products continuously. All checks and findings are recorded to CyberFort Blockchain™.
Hover or focus a band to see what it guarantees.
Inputs and outputs
- IN-01Commander intent
- IN-02Task organisation
- IN-03Constraints, restraints and assumptions
- OUT-01Task decomposition trees
- OUT-02Dependency and sequencing checks
- OUT-03Contradiction and gap reports
Position in the mesh
- Consumes
- Publishes
- CommanderNLCommand interface
- SC-PLAN-01Operational Planning Agent
What it does not do
It does not generate plans of its own. It is the sceptic in the mesh, and it stays that way.
The same four gates as every agent
An unregistered or unapproved model tag is refused before inference, not flagged afterwards.
Every action is scoped to a declared role. Scope is checked at the engine, not in the prompt.
New or changed models run under supervision until their evaluation evidence is accepted.
Irreversible decisions require a signed human authorisation recorded with the decision.
The agents differ in what they reason about. They do not differ in what they are allowed to do.
Deployment
In-country, accredited cloud, under your jurisdiction and operational control.
Designed for deployment up to NATO SECRET. Accreditation is per deployment.
Runs disconnected. Inference, agents and the decision record are all local.
Ready to see SynapseCommand® in action?
Briefings available under NDA. Classified briefings arranged through cleared channels.