Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Peiiii/nextclaw --skill codex-narp-runtimegit clone --depth 1 https://github.com/Peiiii/nextclawWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/peiiii/nextclaw/codex-narp-runtime)<a href="https://agentmods.dev/skills/peiiii/nextclaw/codex-narp-runtime"><img src="https://agentmods.dev/badge/skills/peiiii/nextclaw/codex-narp-runtime/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/peiiii/nextclaw/codex-narp-runtime"><img src="https://agentmods.dev/badge/skills/peiiii/nextclaw/codex-narp-runtime.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00059 | $0.03091 |
| Opus 5 | $0.00030 | $0.01545 |
| Sonnet 5 | $0.00012 | $0.00618 |
| Haiku 4.5 | $0.00006 | $0.00309 |
Grade A, and why
codex-narp-runtime scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 236 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Codex NARP Runtime
使用这个 skill 时,目标不是让用户自己准备 PATH 或手改配置,而是把 Codex 接成 NextClaw 统一 runtime registry 里的正式会话类型。
首选产品路径固定为:
- runtime entry id:
codex - runtime entry label:
Codex - runtime type:
narp-stdio - wire dialect:
acp - runtime launcher:
nextclaw-codex-narp - route ownership:
NextClaw -> RuntimeRoute(model/apiBase/apiKey/headers) -> Codex NARP wrapper -> Codex SDK/CLI
不要把 Codex 做成核心系统里的 provider 特判。NextClaw 核心、kernel、service 只感知 narp-stdio,不感知 codex。
安装边界
始终区分三层:
- 这个 NextClaw marketplace skill:负责用户接入流程、检查、修复和冒烟。
- Codex NARP wrapper:
@nextclaw/nextclaw-narp-runtime-codex-sdk,提供nextclaw-codex-narp。 - Codex SDK/CLI:由 wrapper 调用,真实执行 agent loop。
安装这个 skill 不等于 Codex runtime 已经可用。skill 必须自己检测 launcher,缺失时安装或修复 wrapper,并生成 NextClaw 可稳定调用的绝对路径 shim。
这个 Skill 负责什么
- 解释 Codex 通过 NARP stdio 接入的正确心智模型。
- 发现、安装、显式更新或修复
nextclaw-codex-narp。 - 在
${NEXTCLAW_HOME:-~/.nextclaw}/bin下生成稳定 shim。 - 写入或修复
agents.runtimes.entries.codex。 - 运行 runtime probe。
- 运行真实 Codex 首条消息冒烟;需要能力验收时继续覆盖工具调用和思考事件。
- 对常见失败给出有界诊断,不把问题推给用户自己猜。
禁止事项
- 不要告诉用户“先把
nextclaw-codex-narp放到 PATH 里”作为完成条件。 - 不要要求用户手工编辑
config.json,除非当前 agent 没有文件写权限。 - 不要修改通用 NARP stdio client 来识别 Codex。
- 不要在 core/kernel/service 注入 Codex 默认 entry 或 provider 分支。
- 用户没有明确要求“更新”“升级”或指定目标版本时,不要查询最新版或改动一个已经可用的 wrapper。
- 显式更新不能用“现有 launcher 能运行”或“源码仓库里有 dist”代替版本升级。
- 没有 probe 和真实模型回复时,不要声称已经完成接入。
Setup 流程
当用户要求“接入 Codex”“安装 Codex runtime”“doctor Codex runtime”或类似目标时,按下面顺序执行。
1. 确认工作目录和数据目录
先确定 NextClaw 数据目录:
NEXTCLAW_HOME="${NEXTCLAW_HOME:-$HOME/.nextclaw}"
需要写入:
$NEXTCLAW_HOME/bin/nextclaw-codex-narp$NEXTCLAW_HOME/config.json
如果这是首次接入,先说明 Codex runtime 会读取本机工作区并可能执行本地工具;涉及删除、发送、提交、发布等外部可见动作时仍需用户确认。
2. 解析可用 launcher
按优先级解析真实 launcher:
- 如果当前在 NextClaw 源码仓库,并且存在已构建文件:
packages/extensions/nextclaw-narp-runtime-codex-sdk/dist/controllers/codex-narp.controller.js - 如果源码存在但 dist 不存在,先构建:
pnpm --filter @nextclaw/nextclaw-narp-runtime-codex-sdk build - 如果系统已有
nextclaw-codex-narp,可作为真实 launcher 来源。 - 否则安装 wrapper 包到 NextClaw 管理目录:
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 236 lines · 59 tokens per session scan A 60d25e951569
codex-narp-runtime is a skill published in the GitHub repository Peiiii/nextclaw (256 stars, last pushed today), licensed MIT. It adds 59 tokens to every session and 3,091 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
openclaw-github-dedupe
Investigate a cluster of GitHub issues and PRs, determine canonical candidates, post duplicate/related status, preserve contributor credit, and execute cleanup actions. Supports autonomous mode for provided-link-only closeout, merge/fix follow-through, changelog, and post-merge issue/PR cleanup.
openclaw-pr-batch-sweep
Select, review, repair, validate, and land batches of up to 20 low-risk OpenClaw contributor pull requests using Vincent's maintainer preferences and bounded sub-agent lanes. Use for "next 20", broad contributor PR sweeps, merge-candidate mining, or continued PR-batch work where drafts, maintainer work, trivial…
tmux-lane-orchestrator
Manage one tmux agent lane from its matching ops pane, inspect live pane state and Codex logs on cold start, and produce concise manager summaries for OpenClaw and adjacent project work.
ghcrawl-cluster-operator
Use when inspecting a ghcrawl SQLite store, pulling GitHub issue/PR data, refreshing summaries, embeddings, and clusters, or extracting one cluster and its evidence through the ghcrawl CLI.
opik-optimizer
Optimize LLM prompts, tools, and agents in Opik using standardized optimizer workflows (prompt optimization, tool optimization, and parameter tuning), dataset/metric wiring, and result interpretation.
org-branch-cleanup
Audit and safely prune stale branches across a GitHub organization with immutable snapshots, conservative merged-PR classification, live SHA/protection/open-PR revalidation, resumable deletion ledgers, and post-delete verification. Use when a maintainer asks to clean up old, dead, merged, bot-created, or abandoned…