← 返回 Skills 市场
My Research Lab
作者
iriswong31
· GitHub ↗
· v1.8.0
· MIT-0
165
总下载
0
收藏
0
当前安装
10
版本数
在 OpenClaw 中安装
/install my-research-lab
功能描述
My Research Lab 我的研究院 — Your personal research lab for any topic you care about. Auto-running with cron scheduling, self-evolving. From discovery to brainsto...
安全使用建议
Before installing: 1) Inspect the contents of ~/.codebuddy/memory and ~/.codebuddy/automations to confirm they don't hold unrelated secrets or tokens (backup first). 2) If you must provide SMTP creds, use an app-specific password or dedicated mailbox with limited privileges and monitor/follow strong revocation steps. 3) Be prepared that the skill will create and modify cron jobs — back up your crontab and review any newly created job entries before allowing them to run autonomously. 4) Ask the publisher (or request the exact prompts/paths the skill will write) to clarify which files it will read/write and where it will store logs. 5) If you have sensitive data in shared memory used by other skills, avoid granting this skill access until you can scope its storage to an isolated directory. If you are uncomfortable with autonomous auto-repair or background scheduling, do not install or run it in an unprivileged/test environment first.
能力评估
Purpose & Capability
The skill claims to be an autonomous, cron-driven research lab and requests SMTP_USER/SMTP_PASS and two config paths (~/.codebuddy/memory and ~/.codebuddy/automations). SMTP credentials are reasonable for an email-pushing research-digest feature. The config paths are plausible for storing memory/automations, but the prefix '.codebuddy' is external to the skill name ('my-research-lab') and suggests it will read/write data belonging to a different subsystem; that deserves scrutiny.
Instruction Scope
SKILL.md instructs the agent to read and update user memory, create and modify cron tasks, perform web searches/fetches, access logs and error diagnostics, store archives locally, and send HTML emails via SMTP. Many of these actions are coherent with the stated purpose, but the instructions are broad and include automated repair actions ('must recreate cron if missing', 'auto-fix channels') and vague reads of '错误日志' ('read and diagnose last execution error logs') without clear path restrictions — this grants the skill wide discretionary access to system state and files beyond a narrowly scoped application folder.
Install Mechanism
Instruction-only skill with no install spec or downloaded code. That limits supply-chain risk because nothing new is written to disk by an installer. The runtime instructions themselves (create cron jobs, write files) are the main risk surface rather than an install mechanism.
Credentials
Only SMTP_USER and SMTP_PASS are declared as required env vars, which fits the email-sending functionality. However the skill requires read/write access to specific config paths (~/.codebuddy/memory and ~/.codebuddy/automations). Those paths may contain unrelated secrets or tokens from other skills (the '.codebuddy' name suggests a shared platform memory). Requesting those paths without explicit scope increases risk of accessing unrelated credentials or private data.
Persistence & Privilege
The skill will create and manage cron jobs and persist archives locally. Creating system scheduled tasks and auto-fixing them is a high-impact capability. Although 'always' is false and autonomous invocation is the platform default, the combination of filesystem access (~/.codebuddy/...), cron modification privileges, and SMTP credentials gives it significant persistent presence and potential to act without frequent user interaction. The SKILL.md promises user confirmations for direction selection, but also instructs '能自己查的不问用户' which pushes for fewer user prompts — a potential scope creep toward more autonomous changes.
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 my-research-lab - 安装完成后，直接呼叫该 Skill 的名称或使用
/my-research-lab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
v1.8.0
v1.8.0: SECURITY FIX - Added metadata.openclaw.requires (env: SMTP_USER/SMTP_PASS, config: memory/automations) to match actual runtime behavior. Added explicit permissions declaration section. Removed 'silent' wording that triggered suspicious pattern detection. All permissions now transparently documented.
v1.7.0
v1.7.0: Cron timeout must be >=15min (fixes 5min timeout killing tasks). Email MUST use HTML format with card layout and clickable links. Summary quality check added (must have data+judgment+info beyond title). Original URL mandatory for every item. Archive links in footer. Auto-retry on push failure. Auto-diagnose when tasks fail.
v1.6.0
v1.6.0: CRITICAL - Cold start now MUST create cron scheduled task (no more fake auto-run). Added channel availability detection before config. Added scheduled task health check on every conversation. Added cloud environment auto-downgrade (no Chrome = web_search + web_fetch only). Content creation is suggestion-only. Hands-on projects use suggest-confirm-authorize flow.
v1.5.0
v1.5.0: Removed CDP scripts (fixes SUSPICIOUS flag), web-access now recommended on-demand with full risk disclosure, hands-on projects use suggest-then-authorize flow, content creation is optional suggestion not mandatory, ultra-minimal onboarding (one question)
v1.4.0
v1.4.0: Ultra-minimal onboarding (one question only), proactive push channel setup (QQ mail/WeChat/etc with step-by-step auth guide), content creation engine (auto-suggest weekly articles for WeChat/Xiaohongshu/email), auto-update check on each run, --force install note for CDP scripts
v1.3.1
v1.3.1: Fixed display name to English, added bilingual CN/EN description and title
v1.3.0
v1.3.0: Rebranded to '我的研究院'. Added 5-tier search tool matrix, 3 self-evolution designs (source curation, quality scoring >= 3.5, self-improving feedback), Skill security policy (100/3 rule + dual audit), auto-run heartbeat mechanism, content quality scoring system
v1.2.0
v1.2.0: Added browser CDP automation (web-access integration), 3-tier web capability (search -> fetch -> browser), YouTube/video support, parallel sub-agent research, site experience accumulation
v1.1.0
v1.1.0: Built-in web access capabilities (web_search + web_fetch), enhanced source collection with 15+ domain sources, improved brainstorm engine with deep web research, optimized output templates
v1.0.0
Initial release: personal research automation engine
元数据
常见问题
My Research Lab 是什么？
My Research Lab 我的研究院 — Your personal research lab for any topic you care about. Auto-running with cron scheduling, self-evolving. From discovery to brainsto... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 165 次。
如何安装 My Research Lab？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install my-research-lab」即可一键安装，无需额外配置。
My Research Lab 是免费的吗？
是的，My Research Lab 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
My Research Lab 支持哪些平台？
My Research Lab 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 My Research Lab？
由 iriswong31（@iriswong31）开发并维护，当前版本 v1.8.0。
推荐 Skills