← 返回 Skills 市场
Update Advisor
作者
kelven.ling
· GitHub ↗
· v1.1.1
· MIT-0
157
总下载
0
收藏
1
当前安装
7
版本数
在 OpenClaw 中安装
/install update-advisor
功能描述
OpenClaw update check and upgrade assistant. Triggers on phrases like "check for updates", "any new version", "is openclaw updated", "run the update", "confi...
安全使用建议
This skill appears to do exactly what it claims: run local checks and, with explicit confirmation, perform an OpenClaw update protected by a user-level recovery watchdog. Before using: (1) run Check mode first to review the changelog analysis and doctor output; (2) confirm you are comfortable letting the skill create a temporary user-level launchd/systemd job to monitor and, if needed, run 'openclaw gateway install' — the job is created under your user and cleaned up after recovery; (3) ensure openclaw, python3, and npm are installed locally; (4) only allow MEMORY.md reads if you explicitly consent to personalized relevance analysis. Review the bundled scripts (they are included and well-tested) if you want to audit exact behavior; there are no hidden network endpoints or undeclared credential usage.
功能分析
Type: OpenClaw Skill
Name: update-advisor
Version: 1.1.1
The update-advisor skill is a utility designed to manage OpenClaw CLI updates safely. It features a 'Check mode' for changelog analysis and risk assessment, and an 'Execute mode' that employs a temporary recovery watchdog (via macOS launchd or Linux systemd) to ensure the Gateway restarts after the update-induced session disconnect. The implementation includes robust safety checks, such as installation ownership verification (Step 0) and mandatory user confirmation gates. The code (scripts/check-update.sh, scripts/arm-recovery-watchdog.sh) is well-structured, includes cleanup logic, and lacks any indicators of data exfiltration or malicious intent.
能力评估
Purpose & Capability
Name/description (update checking + guarded execution) aligns with the actual requirements and actions: it needs openclaw, python3, and npm; it reads local CHANGELOG.md and may call 'npm view' as a fallback; scripts perform changelog parsing, doctor checks, and optional recovery arming — all directly related to updating OpenClaw.
Instruction Scope
SKILL.md confines actions to Check mode (read/report only) and Execute mode (explicit user-confirmed openclaw update + user-level watchdog). It documents MEMORY.md as optional read-only and requires explicit user consent before reading it. Script steps reference only expected files/commands (CHANGELOG, openclaw update status, launchd/systemd), and do not instruct arbitrary file exfiltration or calls to unexpected external endpoints.
Install Mechanism
No install spec (instruction-only). All code is bundled in the skill directory and executed locally; there are no downloads from external URLs or registry installs. This minimizes install-time risk and matches the stated architecture.
Credentials
The skill requires no secrets or cloud credentials. The only network lookup (npm view openclaw version) is documented and justified as a fallback. Scripts accept optional environment overrides (OPENCLAW_BIN, WATCHDOG_BACKEND, etc.) for testing and portability, which is reasonable and declared.
Persistence & Privilege
The skill does create temporary user-level launchd/systemd jobs during Execute mode to arm a recovery watchdog — this is necessary for its recovery behavior but is a persistent side effect while the watchdog runs. always:false and explicit user confirmation gating are present; verify you want the skill to create/clean up user-level service units before granting Execute permission.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install update-advisor - 安装完成后，直接呼叫该 Skill 的名称或使用
/update-advisor触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.1.1
Clarify security metadata and consent boundaries for ClawHub audit: declare required tools, make MEMORY.md access optional/read-only with explicit consent, and document Execute watchdog side effects without changing core functionality.
v1.1.0
Add Gateway-independent Execute recovery watchdog: safer rollback guidance, macOS launchd arming, Linux systemd user-unit arming, bounded recovery retries/timeouts, and mock-only safety tests.
v1.0.4
Harden update recovery flow: prefer OpenClaw update JSON metadata, verify cron scheduling before update, add dry-run checks, safer rollback guidance, stronger ownership and doctor issue detection.
v1.0.3
fix: add Gateway restart fallback in post-update cron verification - isolated session now executes 'openclaw gateway start' if all version-check retries fail, plus exec environment risk documentation
v1.0.2
Fix: remove --yes flag from openclaw update command to resolve Gateway restart failure after updates
v1.0.1
Add implementation architecture note clarifying that execute mode uses OpenClaw built-in cron/exec tools (not shell scripts), MEMORY.md access is read-only for personalization, and all destructive actions require explicit user confirmation.
v1.0.0
Initial release. Safely check and apply OpenClaw updates with changelog analysis, risk rating, multi-user ownership detection, and automatic post-update verification via cron.
元数据
常见问题
Update Advisor 是什么？
OpenClaw update check and upgrade assistant. Triggers on phrases like "check for updates", "any new version", "is openclaw updated", "run the update", "confi... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 157 次。
如何安装 Update Advisor？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install update-advisor」即可一键安装，无需额外配置。
Update Advisor 是免费的吗？
是的，Update Advisor 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Update Advisor 支持哪些平台？
Update Advisor 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Update Advisor？
由 kelven.ling（@lzyling）开发并维护，当前版本 v1.1.1。
推荐 Skills