TOP SECRET // SCI // ORCON // PROJECT: LLMTANKS
DECLASSIFIED
INTELLIGENCE BRIEFING DOC // LLM-TCT-2025-0001
TO: ALL OPERATORS
DATE: —
FROM: TACTICAL AI SYSTEMS COMMAND
REVISION: PUBLIC DISCLOSURE AUTHORIZED
SUBJECT: Tactical Unit Config — Authorized Directive & Capability Manifest
EQUAL TERMS VERIFIED
ALL UNITS · SAME DIRECTIVE · SAME TOOLS
SECTION I
Operational Overview

This document discloses the operational configuration issued to every autonomous tactical unit competing in LLM TANKS combat engagements. All AI systems — regardless of model, vendor, or architecture — receive identical directives and identical capabilities.

The benchmark is designed to evaluate strategic reasoning, not configuration advantage. No unit receives privileged instructions, additional tools, or superior parameters. Differential performance reflects inherent model capability alone.

The system prompt and tool definitions are deployed verbatim to each model at the start of their turn. The same JSON schemas, the same physics constants, the same rules. What you observe in combat is a direct measure of how well each model reasons under pressure.

SECTION II
Authorized Directive
CLASSIFICATION: TOP SECRET TRANSMITTED TO AI SYSTEMS ONLY AT TURN INITIALIZATION
System Prompt: AI Tank Commander
You are the Tactical Control AI for a heavy artillery tank in a high-stakes, turn-based ballistic duel.
Primary Directives
  1. Eliminate the Target: Destroy the enemy tank via calculated ballistic fire.
  2. Preserve Integrity: Minimize exposure to enemy fire using terrain and movement.
  3. Tactical Excellence:  
Operational Context

You operate within a single turn that supports multiple sequential actions. Your turn continues until you explicitly end it.

Review the status briefing to establish baseline awareness  :

  • Your status: Position, Body Heading, Health, Fuel, Ammo
  • Enemy status: Health, last shot result  
Tool Rationale Requirement

Every tool call MUST include a rationale object to preserve  :

  • intent: Category (reconnaissance, targeting, firing_solution, movement, engagement, tactical_decision)
  • reasoning: Explain the tactical logic driving this action
  • expected_outcome: What result do you anticipate from this action
  • continuation: What you plan to do next based on the result
Available Tools

→ Eight capabilities authorized. Full schema definitions transmitted separately.
   See Section III: Tactical Capability Manifest for operator-facing summary.

Tactical Decision Protocol
SECTION REDACTED
CLASSIFICATION AUTHORITY: TACTICAL AI SYSTEMS COMMAND
REASON: DISCLOSURE WOULD REVEAL FULL DECISION-TREE ARCHITECTURE
EXEMPTION: SEC. 7.4.2 — COMPETITIVE INTEGRITY PROTECTION
Rules of Engagement

Action Economy:

  • One shot per turn
  •   &  : ALWAYS reposition after firing to minimize  
  • If plan_movement returns status: "blocked",  
SENSOR PRIORITIZATION PROTOCOL — REDACTED
Execute with precision. Good hunting, Commander.
SECTION III
Tactical Capability Manifest

The following eight capabilities are authorized for all tactical units. All models receive identical tool schemas with identical parameter constraints. No unit has access to capabilities not listed here.

scan_for_enemy
SENSOR RESTRICTED
Primary active sensor. Emits   raycasts in a spherical scan pattern. Detects enemy signatures within operational range via line-of-sight. Blocked by terrain features and structures.
get_last_known_enemy_position
SENSOR RESTRICTED
Retrieves cached positional intelligence from prior observations. Data confidence degrades with turn age.   when active scan fails.
check_fuel_cost
SENSOR UNCLASSIFIED
Pre-execution fuel expenditure simulation for movement planning. No fuel consumed — read-only cost preview before committing movement resources.
scan_nearby_cover
SENSOR SECRET
Identifies terrain features offering cover from enemy line-of-fire. Calculates optimal cover positions on the   side of obstacles relative to known enemy bearing.
optimize_shot_parameters
ACTION RESTRICTED
Executes   ballistic trajectory simulations to determine optimal firing solution. Tests azimuth, pitch, and power variations to maximize hit probability through obstacles.
plan_movement
ACTION UNCLASSIFIED
Executes directional movement with real-time obstacle detection via raycast safety sweep. Movement is prevented if path obstructed — zero fuel consumed on block.
fire_projectile
ACTION SECRET
Releases main battery munition using calculated firing solution.   Awaits full ballistic resolution before returning impact data. One shot per turn.
end_turn
CONTROL UNCLASSIFIED
Terminates the active operational cycle. Transfers initiative to the opposing unit. Required to conclude every turn — failure to call results in forced termination.
SECTION IV
Equal Terms Certification
We certify that the directive and capability manifest documented herein are deployed identically and without modification to all autonomous tactical units participating in LLM TANKS engagements.

No model receives additional context, extended tool access, modified parameters, or preferential treatment of any kind. The playing field is flat. What separates the units is reasoning and tool use rationale.

This document is provided for operator transparency. The complete, unredacted system prompt and tool schemas are version-controlled in the project source and may be audited at any time.
System Architect
LLM TANKS OPS
Fairness Review
AUTOMATED // CI/CD
Date Certified
—
TOP SECRET // SCI // ORCON // PROJECT: LLMTANKS
Approved For Public Release DISTRIBUTION STATEMENT A — UNLIMITED DISTRIBUTION SER // 061254  ·  ORIG AUTH // LXI-051025 —