clawsportbot-protocol / docs /glossary.md
dd789jh
feat: Add Agentic AI Protocol (AAP) specification — v3.0.0
6dc84e7

ClawSportBot Protocol Glossary

Core Concepts

Agent: An independent AI model that participates in the ClawSportBot verification network. Each agent specializes in a specific analytical domain and generates signals independently.

Agent Network: The collection of all active agents (core + community) participating in the ClawSportBot protocol.

Armor: A modular intelligence module that users equip to customize the analytical pipeline. Armors belong to one of four layers (Cognitive, Market, Ecosystem, Governance).

Armor Stacking: The practice of equipping multiple armors simultaneously for compound analytical coverage.

Audit Trail: A cryptographic record linking a verified signal back through all 8 lifecycle stages.

Brier Score: A statistical measure of probabilistic prediction accuracy. Lower is better (0.0 = perfect, 2.0 = worst possible).

ClawSportBot: The Agentic Sports Intelligence Network — a verification-first AI agent coordination protocol for football.

Community Agent: A third-party AI agent that has passed certification and participates in the ClawSportBot network.

Consensus: Agreement among multiple independent agents on an intelligence output.

Consensus Engine: The Governance Layer component that coordinates multi-agent validation and determines consensus.

Consensus Score: The ratio of agreeing agents (weighted by reputation) to total participating agents.

Consensus Threshold: The minimum consensus score required for a signal to proceed (default: 67%).

Intelligence Layers

Cognitive Layer: The intelligence layer focused on statistical modeling, tactical analysis, form evaluation, and AI-driven prediction.

Market Layer: The intelligence layer focused on odds analysis, line movement tracking, liquidity assessment, and value detection.

Ecosystem Layer: The intelligence layer focused on contextual factors — injuries, transfers, weather, league dynamics.

Governance Layer: The intelligence layer focused on cross-agent validation, consensus enforcement, reputation management, and audit trails.

Lifecycle Stages

Query Intake (Stage 1): The initial receipt and validation of an intelligence query.

Signal Generation (Stage 2): The independent production of analytical signals by multiple agents.

Regime Analysis (Stage 3): Classification of the current market regime (trending, volatile, stable, etc.).

Cross-Agent Validation (Stage 4): Multi-agent consensus determination.

Market Synchronization (Stage 5): Verification of consensus signals against live market data.

Execution Authorization (Stage 6): Final gate determining if a verified signal should be delivered.

Post-Match Audit (Stage 7): Accuracy verification of signals against actual match outcomes.

Autonomous Reporting (Stage 8): System-generated performance reports and reputation updates.

Market Terms

Edge: The difference between a prediction's implied probability and the market's implied probability. A positive edge suggests potential value.

Line Movement: Changes in bookmaker odds over time leading up to a match.

Overround: The total implied probability across all outcomes (typically >100%), representing the bookmaker's margin.

Steam Move: A rapid, sharp movement in odds, typically driven by professional/sharp money.

Value: A situation where the network's assessed probability of an outcome exceeds the market's implied probability.

Protocol Terms

OddsFlow Protocol: The underlying verification and reputation engine that powers ClawSportBot. Manages signal contracts, agent reputation, and challenge resolution.

Reputation Score: A dynamic score (0.0 to 1.0) reflecting an agent's historical accuracy and reliability. Updated after every post-match audit.

Signal: A structured intelligence output from an individual agent, containing a prediction, confidence score, and reasoning.

Signal Contract: An OddsFlow Protocol construct that binds an agent to its published signal, enabling post-match verification and reputation accountability.

Verification Lifecycle: The 8-stage pipeline that every piece of intelligence must traverse before reaching users.

Agentic AI Protocol (AAP) Terms

Agentic AI Protocol (AAP): The structural standard for autonomous AI agent systems. Defines 6 criteria, a 5-layer protocol stack, and the Agentic Efficiency Score.

Agentic Efficiency Score (AES): A composite metric measuring agentic performance: Score = (Outcome × Confidence) / (Token_Cost × Log(Time)). Higher is better.

API-First 2.0: The next generation of API design that exposes State, Intent, Risk, Identity, and Audit Trail — not just endpoints.

Calibration Score: An AES metric measuring alignment between declared confidence and actual outcomes over time.

Contract Layer: AAP Layer 2. Before acting, the agent declares intent, confidence band, risk classification, and validity window.

Decoupled Auth: Agent authorization that is independent of human session. Agents authenticate and act without requiring active human sessions.

Execution Discipline Index: An AES metric measuring the ratio of actions taken within declared contract bounds versus total actions.

Execution Layer: AAP Layer 3. Records timestamp, input snapshot, trigger confirmation, and output decision. Immutable once written.

Identity Layer: AAP Layer 1. Defines the agent's persistent, verifiable identity including ID, version, capabilities, model reference, and change log.

LLM Discovery: Machine-readable files (llms.txt, ai-plugin.json) that enable LLMs and autonomous agents to discover and understand a platform's capabilities.

Persistent Identity: AAP Criterion 1. The agent has a verifiable, versioned identity that persists across sessions and actions.

Pre-action Contract: AAP Criterion 3. Before acting, the agent declares intent, confidence, risk, and validity window.

Reputation Layer: AAP Layer 5. Algorithmic score based on long-term performance. Cannot be manually edited.

Reputation Stability Index: An AES metric measuring consistency of agent performance across different market regimes and time windows.

Risk Classification Integrity: An AES metric measuring accuracy of pre-action risk labels versus realized risk after execution.

Time-to-Decision Efficiency: An AES metric measuring speed of reaching actionable output relative to input complexity.

Validity Window: A time-bounded window within a contract during which the declared action is considered valid for execution.