openclaw

An agent skill by daymade, from daymade/claude-code-skills. Tags: ai, catalog, claude-code, configuration, validation.

What it does

Manage OpenClaw (龙虾) instance configurations. Use whenever the user wants to audit, diff, copy, add-model, list, or switch models in an openclaw.json file, or when they mention lobsters, 虾, 甲虾, 乙虾, DeepSeek patch, default model, model aliases, or OpenClaw config validation.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add daymade/claude-code-skills --skill openclaw

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/daymade/claude-code-skills
cp -r claude-code-skills/openclaw ~/.claude/skills/openclaw

Safety box score

Grade D, elevated risk: 46 of 100 risk points, rated 2026-10-23. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

OpenClaw is a local config-management skill that reads and rewrites `openclaw.json`-style files, can copy provider definitions between configs, and can restart the gateway service. The main risks are service control, writing user config files/backups in home directories, and moving provider credentials between configs.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	2 (moderate)	The scripts shell out to external binaries, including `openclaw` and `systemctl`, via `subprocess.run(...)` in `restart_gateway()`.
Network access	1 (low)	The shipped operations guide explicitly tells the agent to run `npm view openclaw dist-tags --json` and `npm pack <spec> --silent && tar xzf *.tgz`, which reaches npm and downloads package tarballs.
Filesystem reach	2 (moderate)	It reads and writes user config files in `~/workspace/.force/openclaw/openclaw.json`, `~/.kimi_openclaw/openclaw.json`, and `~/.openclaw/openclaw.json`, and `backup_config()` writes `config-backups/<stem>-<utc-timestamp>.json` beside them.
Secrets & credentials	2 (moderate)	Copy operations can duplicate whole provider objects from one config to another (`copy.deepcopy(source_providers[args.provider])`), which in this config schema includes sensitive fields such as API keys.
Privilege & persistence	2 (moderate)	`restart_gateway()` tries both `systemctl --user restart openclaw-gateway` and `systemctl restart openclaw-gateway`, so the skill can control a user or system service.
Instruction hijack surface	0 (none)	No file tells the agent to ignore its rules, follow remote instructions, hide actions, or otherwise alter its instruction hierarchy.
Irreversible actions	2 (moderate)	`copy`, `add-model`, and `switch` make persistent config changes and can restart the gateway with `--restart`; `switch` directly rewrites `agents.defaults.model`.
Transparency	0 (none)	All shipped content is plain text Python/Markdown/JSON; there are no binaries, minified blobs, or decoded-at-runtime payloads.

Before installing

Read `scripts/openclaw_config.py`: it writes backups under `config-backups/` and can restart services.
Check whether the configs you plan to touch contain `apiKey` or other secrets before using `copy`/`add-model`.
If you do not want service control, avoid `--restart` and test on a throwaway config path first.
Review `references/openclaw_operations.md` if outbound npm access is disallowed in your environment.

Rated 2026-10-23 by an automated review of 14 files from the skill's repository.

Source

Repository: daymade/claude-code-skills (all skills from this repository)
Path: openclaw/SKILL.md
Branch: main
Updated: 2026-09-20

openclaw-model-switch — Switch or repair the model configuration of an OpenClaw instance (e.g. Kimi k2p6 → k3): change the default model, add model definitions.
skill-governance — Govern the real Claude Code and Codex Skill surface without losing cold capability.
agents-pay — Use when THIS agent needs to pay for x402-protected content at runtime: hitting a paywall mid-task, settling it via AgentCore Payments.
check-agent-compatibility
principle-encode-lessons-in-structure — Apply when you catch yourself writing the same instruction a second time, or notice a recurring correction.
review-plugin-submission — Audit a Cursor plugin for marketplace readiness. Use when validating manifests, component metadata, discovery paths, and submission quality before publishing.