HUMAN.md

A document describing who a person is in the context of the work — their role, what they own, what they defer, their voice, their strong suits, and how to operate with them. It is read at session start so an agent works with the person rather than at them.

HUMAN.md is both a file convention in this repo (HUMAN.md at the repo root, describing John) and a foundational Dotabase in the product (slug human-md), so any user can author their own and have agents read it. It answers the questions a competent new colleague would need answered on day one and that no amount of world knowledge supplies.

The repo's HUMAN.md is explicit about its purpose: John is not a developer, does not want to be in the build loop, owns product/IA/taxonomy/pricing decisions, and defers code architecture and test strategy to the agent.

Usage examples

• "Read AGENT-GOVERNANCE.md, DESIGN.md, and HUMAN.md." (step 1 of the Claude Code session protocol)
• "HUMAN.md says don't escalate with a stack trace — escalate at a product/behavior level."
• "Your HUMAN.md documents — who you are, in the context of the work." (the HUMAN.md Dotabase description)

Source

HUMAN.md (repo root), whose subtitle reads "Who John is, the role he's playing on this project, and how to operate with him. Read alongside CLAUDE.md at session start." Listed as a foundational Dotabase at AGENT-GOVERNANCE.md § Architecture; spec row SPEC-160 (DS_HUMAN-MD).