simnova

openclinxr-openclaw-style

Use when operating OpenClinXR's OpenClaw-style / OpenClaw-inspired repo-native workflow in Codex, including role consultation, run-next, leases, hooks, and canonical state updates.

simnova 0 Updated 2mo ago
GitHub

Install

npx skillscat add simnova/openclinxr/openclinxr-openclaw-style

Install via the SkillsCat registry.

SKILL.md

OpenClinXR OpenClaw-Style Bridge

This skill packages the repo-native OpenClaw-style / OpenClaw-inspired workflow for Codex. It is not an external OpenClaw runtime.

Start

  1. Read AGENTS.md.
  2. Read the snapshot heads of PROJECT_STATUS.md and docs/openclinxr/worker-backlog-and-validation-matrix.md (legacy PROJECT_COORDINATION_INDEX.md / AUTONOMOUS_WORK_PLAN.md ledgers were purged 2026-08-05).
  3. Use pnpm openclaw:run-next to select a real approved slice when continuation is needed.
  4. Acquire a lease before edits with pnpm openclaw:lease -- acquire --owner <owner> --slice "<slice>" --ttl-minutes 60.
  5. Use Codex custom agents from .codex/agents/*.toml only when explicitly asked for subagents/delegation or when the repo instructions require role consultation.

Guardrails

  • Treat agents/**/charter.md, agents/**/memory.md, and .agent-factory/memory-index.json as file-backed role memory.
  • Keep canonical state in the repo files named by AGENTS.md, not chat-only summaries.
  • Run focused verification for touched code and pnpm agent:alignment && pnpm docs:drift-check after coordination/OpenClaw-style changes.

Categories