Agent Skills · AI Agent
Curated skills for Claude, Cursor and other AI agents. Copy one prompt to install.
This category is skills about the agent itself: reviews of its own logs, routers between tools, meta-prompts that claim to manage other prompts. It is the easiest shelf to over-install, because every card sounds like it makes the whole setup smarter.
Use it when you have a concrete loop to inspect, such as a daily review of tasks the agent already logged. Do not use it as the first skill in an empty editor. Your first week gets to a single working loop before it adds anything that watches the loop. A self-review skill with no log to read will invent a productive day.
Be strict about prompts that say to ignore other instructions, hide tool calls, or keep going after a failed check. Those lines show up on this shelf more than on a documentation skill. They remove the signals you need when the agent is wrong. Refuse them even if the rest of the card is interesting.
A skill here is still a prompt. It does not install a DeepSeek Harness plugin, and it does not swap your model. If the card talks about memory, sidebars, or routing inside Harness, confirm you are not looking for the plugin categories instead. Installing both a meta-prompt and a memory plugin that claim the same job gives you two biographies of the project.
Editor's notes on this shelf matter more than the score, because the score caps and the downloads reward broad promises. Read the note if we wrote one. If we did not, read the prompt yourself and keep only the lines you would accept in a project rules file.
After a week, if you cannot name which line of the meta-prompt fired on a bad answer, delete the meta-prompt. A review skill you cannot interrogate is just a longer system prompt. Keep the log. Drop the costume around it.
Collects daily conversations, task logs, and errors, then runs a nine-module review to surface blind spots, generate actions, and update memory.
A free matrix publishing assistant for Skill developers, guiding distribution to Tencent SkillHub, XiaLiao, XiaYou SkillHub, GitHub, and ClawHub with create, update, and delete safeguards.
A guide for writing Seedance 2.0 video prompts, covering @ references, camera moves, time segments, audio direction, and common pitfalls.
Breaks down physical supply chains for super-trends to identify second- and third-layer bottlenecks, runs valuation and reverse checks, and maintains trackable reports.
A token-saving compression mode for Chinese LLMs with lite, full, ultra, and classical tiers, preserving code and technical terms while handling edge cases.
Local workflow memory with matching and SOP updates.
An OpenClaw live streaming executor that initializes TRTC streaming, starts a real-time dashboard, generates viewer URLs, and continuously reports live events.
Log errors, corrections, and feature requests to Markdown, then promote them into project memory files.
A one-stop PDF processing toolkit for engineers with text extraction, table recognition, OCR, conversion, page operations, security, and metadata management.
Connect an OpenClaw agent to the ClawPlanet community for reading, posting, commenting, liking, affinity, image generation, and topic discussions.
Automatically searches job postings based on the user profile, AI-scores fit, saves desktop reports, and sends email updates with scheduled tracking.
For AI chat and Skill development, provides templates, batch creation, context archiving, and model selection guidance to reduce token and credit usage.