Plugin Claude Code
Autonomous development loop: plan, ship, run codex adversarial-review, and iterate on findings until approved or threshold/iteration cap is reached. v0.3.2 adds --from-pr [ ] — a first-class entry path for mid-development PR handoff. When plan/implement/ship were done outside clodex, --from-pr bootstraps .clod.
Instructions file
Claude Code instructions for lukaskucinski/clodex, covering claude.md — clodex project guidance, what clodex is, architecture: the four-layer stack, why the user kept being right and source of truth.
Agent
Triage codex adversarial-review findings, form a minimal fix plan, implement the fixes on the current feature branch, run local verification, commit, and push. Designed to be dispatched once per clodex review iteration with the codex JSON output as input. Returns a structured summary the orchestrator uses to update…
Agent
Generate a tailored /codex:adversarial-review invocation for the current feature branch, on behalf of a clodex loop iteration. Wraps the existing /codex-focus skill but returns ONLY the captured invocation line — no preamble, fences, or instructions. Use this when the clodex orchestrator needs a clean command string…
Command
Autonomous dev loop: plan → branch → implement → ship PR → /codex:adversarial-review → fix → repeat. Halts on approve, threshold-satisfied, or max-iter hit. Never auto-merges, never applies migrations. FLAGS --max-iter N iteration cap (default: 5) --threshold X highest acceptable severity…
Skill Claude CodeCodex
Autonomous dev loop: plan → branch → implement → ship PR → /codex:adversarial-review → fix → repeat. Halts on approve, threshold-satisfied, or max-iter hit. Never auto-merges, never applies migrations. Resumable + extensible across sessions via .clodex/state.json (v2 schema). FLAGS --max-iter N iteration cap (default…