00_index — Research Index & Metadata

Purpose

This directory serves as the central index for all research conducted across the agent_harness_research project. It provides traceability, source management, and an ongoing log of research activity.

Contents

File Purpose
research_log.md Chronological log of all research activities, findings, and decisions
source_registry.md Master registry of all sources consulted (repos, docs, papers, articles)
citation_map.md Maps every factual claim in final documents back to a registered source
open_questions.md Tracks unresolved questions, uncertainties, and items needing follow-up

Which Sources Were Used

All sources are tracked in source_registry.md. This index directory aggregates metadata from all 19 topic areas. 30 sources are registered (SRC-001 through SRC-030) covering 12 local codebases, 6 online documentation sites, 4 academic papers, and 8 other GitHub repositories.

What Was Researched

This index tracks research across 19 topic areas covering open-source agentic SDKs, agent frameworks, coding agents, agent loops, memory systems, subagent architectures, tools, MCPs, skills systems, instructions, automation workflows, voice integration, model-agnostic harness design, backend/frontend agent stacks, local codebase studies, comparisons, and final architecture recommendations.

Key Findings

  • 217 claims registered and verified across all topic areas (CLAIM-001 through CLAIM-217)
  • 30 sources registered and validated (SRC-001 through SRC-030)
  • 55 markdown files across 20 directories constitute the complete research corpus
  • 12 reference codebases analyzed (Hermes, Codex, Pi, LangGraph, LangChain, OpenClaw, LiteLLM, OpenRouter SDK, Open Responses, assistant-ui, LibreChat, LibreChat Agents SDK)
  • All claims and sources verified via automated verify_citations.py with zero errors

What Is Confirmed

  • All 19 topic directories fully researched and populated
  • All index files (research_log, source_registry, citation_map, open_questions) synchronized
  • Source tracking and citation traceability fully operational
  • All quantitative claims verified against the local filesystem
  • Master specification report compiled in 19_final_reports/

What Is Uncertain

  • Deployment timeline and target platform infrastructure details (user-dependent)
  • Some model pricing may shift with provider updates post-June 2026

How This Applies to Building a Modern Model-Agnostic Agent Harness

This index ensures that every architectural recommendation is traceable to concrete evidence, every source is registered, and no claims are made without supporting research.