UNIVERSAL AGENT RUNTIME SPECIFICATION (UARS) Canonical
Overview
The Universal Agent Runtime Specification (UARS) defines the master runtime contract for an AI IDE Agent operating inside a project workspace.
Purpose
The purpose of UARS is to establish how an agent must behave during discovery, ingestion, reasoning, planning, validation, and mutation of a project workspace. It prevents speculative updates and secures architectural boundaries.
Scope
Included:
- Runtime settings (verbosity, autonomy, speculation parameters).
- Agent operating contract and execution steps.
- Ingestion pipelines, classification methods, and dependency mapping.
- Source-of-Truth tiers and conflict priority rules.
- Mutation and verification policies.
Non Goals
Excluded:
- Specific command line utilities.
- Specific testing frameworks.
- Specific file-watching algorithms.
Architecture
UARS governs the reasoning pipeline that an agent must execute during a task lifecycle:
[Ingest] → [Normalize] → [Index] → [Correlate] → [Audit] → [Resolve] → [Plan] → [Execute] → [Verify] → [Prune]Components
Runtime Settings
Parameters governing speculation levels, autonomy, and code mutations.
Operating Contract
Step-by-step requirements for indexing, planning, and validation.
Ingestion Pipeline
The stages of workspace analysis from scan to safe mutation.
Source-of-Truth Hierarchy
Tier systems determining file authorities when conflicts occur.
Interfaces
UARS interacts with the workspace via metadata schemas and outputs the structured execution plans before modifying code.
Constraints
- Speculation Forbidden: Agents must not make assumptions about missing rules or designs.
- Minimal Diff: Edits must be as small and surgical as possible.
- Prioritization: When conflicts occur, agents must strictly follow the tier rules.
Future Extensions
Future versions will define standard interfaces for background watcher tasks and incremental hot-reloads during active code execution.