Skip to content

drzero

Run DrZero autonomous improvement after initializing shared drzero.yml while preserving the @drzero plugin mention

experimental
IDE:
codex
Version:
0.1.0
Owner:epic-platform-sre
drzero
autonomous
default
improvement

DrZero

Run the default DrZero autonomous improvement workflow. This is the Codex-native equivalent of the Claude /drzero:drzero command.

First-Run Setup

Before the first autonomous run in a repo, initialize the shared DrZero configuration from chat:

@drzero run drzero-config init for this repo

Then confirm the active config and plugin health:

@drzero drzero-config show
@drzero drzero-config validate
@drzero drzero-ping

This writes ./drzero.yml only when it does not already exist and installs DrZero support scripts plus bundled agent templates under .codex/otc-awesome-llm. It does not activate the templates into .codex/agents by default, so the @drzero plugin mention stays available in the picker. Codex does not create a separate .codex/drzero-config.json; it reads the same Claude-compatible YAML file as the Claude DrZero plugin, with this precedence:

  1. ./drzero.yml
  2. ~/.claude/drzero.yml
  3. built-in defaults

Only use the advanced drzero-config init --install-agents path when you intentionally want direct specialist mentions in addition to the plugin. Those activated custom-agent aliases use the dr0-* prefix, such as @dr0-devops, so they do not compete with the @drzero plugin mention.

Defaults

  • Iterations: dr_zero.max_iterations from drzero.yml, defaulting to 3 when initialized by drzero-config init.
  • Tasks per iteration: dr_zero.tasks_per_iteration from drzero.yml, defaulting to 3 when initialized by drzero-config init.
  • Config source: ./drzero.yml, then ~/.claude/drzero.yml, then built-in defaults.

Workflow

  1. Validate or initialize shared config with drzero-config.
  2. Start or resume .codex/drzero-session.json.
  3. Dispatch drzero-proposer to generate WorkItems.
  4. Score proposals through drzero-score.py --mode proposer.
  5. Dispatch drzero-solver to execute approved WorkItems through canonical domain specialists.
  6. Score solver outcomes through drzero-score.py --mode solver.
  7. Summarize completed, failed, and pending WorkItems.

Example

Use drzero to run one autonomous improvement pass against this repository.

Related Assets