Assistants are execution interfaces
Codex, Claude Code, and other assistants can read context, propose changes, and run tools. Their sessions are useful but tool-specific and temporary.
Answer page
A coding assistant performs work in a session; an agent workflow defines the durable repository contract, evidence, and review process around that work.
Direct answer
An AI coding assistant is the tool or agent that helps explore and change code. A coding-agent workflow is the repository-native system that defines what the work is, constrains it, validates it, records evidence, and hands it off for review across tools and sessions.
Practical guidance
Codex, Claude Code, and other assistants can read context, propose changes, and run tools. Their sessions are useful but tool-specific and temporary.
A workflow keeps intent, scope, validation, evidence, and acceptance in repository files that survive a tool change or a new agent session.
The workflow gives the assistant a bounded task and gives the human reviewer a clear evidence trail. Neither layer replaces the other.
Verified demo evidence
$ day-shift init
Created .day-shift workspace
Next: run day-shift doctorAuthorship and sources
Keep exploring
What belongs in versioned repository artifacts when coding assistants do the implementation work.
Read nextSeparate repeatable workflow checks from model-assisted work and human judgment.
Read nextPlain-language explanation of the Human-Agent-Contract: explicit intent, scope, validation, evidence, and reconciliation for coding-agent work.
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.