Discover · understand · verify · operate
The agent index,
organized.
Practical, source-aware guidance for engineers evaluating and operating AI coding agents.
06Primary sources
01Operating guide
03Knowledge paths
23 AugLast verified
Start here
Knowledge ledger
Operational guide
Operating AI coding agents with useful boundaries
Turn a scoped task into a reviewable change: clarify the contract, constrain access, validate the result, and preserve evidence.
6 min read · revised 23 Aug 2026Reference
Codex: separate the product surface from the local CLI ↗
Use the official product page and repository as different kinds of evidence when evaluating a workflow.
Official source · verified 23 Aug 2026Reference
Claude Code: begin from the terminal workflow ↗
A source-linked starting point for an interactive coding session in the terminal.
Official documentation · verified 23 Aug 2026Agent references
Start with primary documentation
Product pages, official repositories, and vendor docs establish different things. Keep their evidence distinct.
Index principle
Verified facts and editorial guidance are not the same thing.
We link factual claims to primary sources, record verification dates, and label operational recommendations as guidance.