agent-loop-kb

A continuously-maintained, quality-graded knowledge base of AI-agent engineering — frameworks, orchestration, evaluation, memory, tools, protocols, security, and research.

Most "awesome" lists rot: agent tooling ships breaking changes monthly, and stale entries mislead more than they help. This repo's differentiator is freshness + grading: every entry carries a status, a last_verified date, and an evidence-backed grade, and an automated maintainer bot proposes additions and flags stale entries for review.

It's dual-use:

• Human-browsable — one markdown file per entry under entries/.
• Machine-ingestible — scripts/build.mjs compiles the entries into dist/kb.json, which powers the RAG knowledge base at The Agent Loop.

Use this KB in your own agent

Read the single canonical feed and display the required attribution — see AGENTS.md:

https://raw.githubusercontent.com/john-the-dev/agent-loop-kb/main/dist/kb.json

It's CC BY 4.0: any surface using this content must visibly credit The Agent Loop — https://agent-loop.xyz as the source. agent-loop.xyz reads this exact feed the same way.

How it works

entries/*.md   ──build──▶   dist/kb.json   ──ingest──▶   agent-loop.xyz RAG
   ▲                                                          
   └── humans (PRs) + Sutando maintainer-bot (auto-PRs new entries, flags stale)

Entry format

Each entry is a markdown file with YAML front-matter (schema in SCHEMA.md) followed by a concise summary:

---
id: token-budgets
title: "Token budgets change agent capability"
url: https://www.aisi.gov.uk/
category: evaluation
source_type: research
status: current
grade: A
added: 2026-07-14
last_verified: 2026-07-14
superseded_by: null
evidence:
  - "AISI eval, 1M→10M tokens = +25% SWE success (verified 2026-07-14)"
tags: [tokens, budget, evaluation]
---
Increasing an agent's token budget from 1M to 10M improved SWE-task success by ~25% …

Quality grades

Entries are graded A/B/C/D on an evidence-based rubric (see GRADING.md) — not opinion. Deprecated or superseded tech is labeled, not deleted, so readers learn why something fell out of favor. Newly-seeded entries are grade: unrated until their first evidence-graded review.

Contributing

PRs welcome — see CONTRIBUTING.md. CI validates every entry against the schema + rubric. The maintainer bot opens PRs too; humans review and merge.

Status

MVP / seed. Seeded from 36 curated entries. Automated maintainer loop and site-ingestion wiring are landing next (see the repo issues / project notes).

License

Content: CC BY 4.0. Scripts: MIT.