The 10 OpenClaw Skills Worth Installing Right Now
I've tested probably 200 OpenClaw skills over the past year. Most of them are half-finished experiments, clones of other skills, or things that worked on the developer's machine and nowhere else. But some are genuinely good. Here are 10 I keep installed on my own agent, ranked roughly by how often I actually use them.
## 1. MailCraft (email management)
MailCraft is the best email skill and it's not close. It reads threads, understands context, and drafts replies that sound like you wrote them. The key differentiator is that it maintains conversation context across a thread. Cheaper email skills treat every message as isolated, so their replies feel disconnected.
The one downside: initial setup requires connecting your email through ClawCoil, and the permission scopes are broad. Worth it, but read through what you're granting.
## 2. DeepDig (web research)
When I need to research a topic, DeepDig saves me an hour every time. It searches, extracts content from the top results, cross-references them, and produces a structured brief with sources. The output is better than what I'd get from 20 minutes of manual Googling because it actually reads the pages instead of skimming.
It's slower than simpler search skills (10-15 seconds per query vs. 2-3), but the quality difference is massive.
## 3. ThreadWeaver (Slack integration)
ThreadWeaver handles Slack with a level of nuance I didn't expect. It reads channel context, respects thread boundaries, and adjusts tone based on the channel. It won't send a casual message in #incidents or a formal one in #random. You configure tone rules per channel in your SOUL.md.
Fair warning: ThreadWeaver is chatty about requesting permissions. It wants access to all public channels by default. Lock it down to specific channels unless you actually need broad access.
## 4. GitForge (GitHub/GitLab)
GitForge does the boring git work: creating PRs from descriptions, reviewing diffs, updating issues, managing labels. I mostly use it for PR descriptions. I describe what changed in a sentence, and GitForge writes a proper PR description with context, testing notes, and linked issues.
It's not going to replace a human code reviewer. But for mechanical git operations, it's a time saver.
## 5. CalSync (calendar management)
CalSync handles scheduling across Google Calendar and Outlook. The killer feature is conflict detection with context. It doesn't just tell you "you have a conflict." It says "this overlaps with your 1:1 with Sarah, which you've rescheduled twice already. Probably don't move that one."
The scheduling suggestions are hit-or-miss. It tends to over-pack mornings if you let it. I set a rule limiting meetings before 10 AM and it behaves much better.
## 6. SheetSense (spreadsheet queries)
SheetSense connects to Google Sheets and lets you query data in natural language. "What were the top 5 products by revenue last quarter" just works if your sheet is reasonably structured. It handles lookups across sheets, basic calculations, and can write data back.
Where it struggles: messy sheets with merged cells, inconsistent headers, or multiple data tables on one sheet. Clean your data first.
## 7. DocDraft (document generation)
I use DocDraft for first drafts of proposals, specs, and reports. It pulls context from previous documents in your connected storage, which means the output matches your writing style and uses your company's terminology. The first draft usually needs 15-20 minutes of editing, which beats the hour it would take to write from scratch.
Don't use it for anything creative or opinion-heavy. It produces competent but bland prose.
## 8. APIBridge (REST API interaction)
APIBridge lets your agent call arbitrary REST APIs by describing what you want in natural language. Give it an API spec (or just a base URL and it'll try to discover endpoints), and it handles auth, request formatting, and response parsing. I use it to check deployment statuses, query internal services, and pull data from tools that don't have dedicated skills.
It's a power tool. It can do almost anything, which means it can also break things if you're not careful with write operations. Start with read-only API keys.
## 9. SnapSumm (meeting summaries)
SnapSumm connects to your meeting recordings (Zoom, Google Meet, Teams) and generates summaries with action items. The summaries are better than the built-in AI features these platforms offer because SnapSumm has context about your projects and priorities. It highlights decisions that affect your work specifically.
Audio quality matters a lot. Calls with background noise or heavy accents produce noticeably worse summaries.
## 10. TaskForge (project management)
TaskForge integrates with Linear, Jira, and Asana. I mostly use it with Linear. It creates issues from conversations, updates statuses, and can triage incoming bugs based on severity patterns it learns over time. The triage feature is underrated. It correctly categorizes about 80% of incoming issues, which means I only manually sort the ambiguous ones.
## Honorable mentions
**CodeScope** is great for codebase exploration if you work across many repos. **TransLay** handles translation better than raw LLM translation because it maintains glossaries. **DataPipe** connects databases for natural language querying, but it's terrifying to give an AI agent database access, even read-only.
## How I evaluate skills
Three things matter more than anything: does it handle errors gracefully (or does it just fail silently), is it actively maintained (check the commit history, not the version number), and does it request only the permissions it needs. A skill that asks for 15 scopes to do one thing is either poorly designed or collecting data. Either way, skip it.
Install these one at a time. Give each a week of real use before adding the next. Stacking 10 skills at once makes it impossible to tell which one is causing problems when something breaks.