Intent starts as a spec
A registered specification gives planning a durable source of truth. It captures the desired outcome and constraints before the task becomes an implementation request.
Answer page
A Human-Agent-Contract is a chain of versioned repository artifacts from intent through validation evidence and acceptance.
Direct answer
In a real repository, a Human-Agent-Contract is not a policy document. It is the linked set of files that states intent, scoped work, validation, changed paths, and acceptance: a spec, planning artifacts, task definition, implementation summary, and reconciliation.
Practical guidance
A registered specification gives planning a durable source of truth. It captures the desired outcome and constraints before the task becomes an implementation request.
The task definition names the objective, target paths, and validation. It gives a coding agent a clear contract while leaving human judgment available for ambiguity.
The implementation summary and reconciliation show what happened and whether acceptance criteria are met. Those artifacts support code review, handoff, and future changes.
Verified demo evidence
$ day-shift reconciliation build --milestone-overview <milestone-overview>
reconciliation_status: completedAuthorship and sources
Keep exploring
Plain-language explanation of the Human-Agent-Contract: explicit intent, scope, validation, evidence, and reconciliation for coding-agent work.
Read nextWhat belongs in versioned repository artifacts when coding assistants do the implementation work.
Read nextA practical checklist for handing work to a coding agent and receiving reviewable implementation evidence back.
Read nextFind your answer
Search the answer pages, concept, comparisons, demo, and FAQ without leaving the site.
Search by a problem, workflow step, or tool boundary.