humanize-artifact-loop

Humanize Artifact Loop (Non-Code Deliverables)

Use this flow to produce file-based deliverables (documents, designs, flow definitions) through an iterative review loop.

Unlike the RLCR loop, this does not use git-based code review. Review is summary-based only.

Required Sequence

1. Setup

"${CLAUDE_PLUGIN_ROOT}/scripts/setup-artifact-loop.sh" $ARGUMENTS

2. Work Round

For each round:

1. Read current loop prompt from .humanize/artifact-loop/<timestamp>/round-<N>-prompt.md
2. Produce deliverable files
3. Commit changes
4. Write summary to .humanize/artifact-loop/<timestamp>/round-<N>-summary.md
5. Run stop gate:

   bash "${CLAUDE_PLUGIN_ROOT}/scripts/artifact-loop-stop-gate.sh"
   

6. Handle gate result: 0 = done, 10 = blocked, 20 = error

Options

Option	Description	Default
path/to/plan.md	Plan file path	Required
--plan-file <path>	Explicit plan path	-
--max N	Maximum iterations	42
--codex-model MODEL:EFFORT	Codex model and effort	gpt-5.4:high
--codex-timeout SECONDS	Codex timeout	5400
--full-review-round N	Full alignment interval	5
--skip-quiz	Skip plan understanding quiz	false

Cancel

Cancel by removing the active state.md from .humanize/artifact-loop/<timestamp>/.