Choose AI tools
with a second opinion
Original guides and editor's notes for agent skills, DeepSeek Harness plugins, and large language models. We tell you when to skip something, not just what to install.
Behind these pages sits a searchable directory of agent skills, DeepSeek Harness plugins, and large language models. Cards describe the publisher's work; our own writing is marked.
An editorial layer over a large directory
AI Agent Hub indexes agent skills, DeepSeek Harness plugins, and large language models. The cards help you find a name, a publisher, and the upstream source. The guides, editor's notes, comparisons, and glossary are written for this site: when a tool fits, when it does not, and how the ordering on this site actually works.
If you are here to decide, start with a guide rather than a search box. Search, sort, and page numbers remain available, but they stay out of the crawl so the pages with a full explanation are the ones search engines spend time on.
Read the guides · How we handle third-party text · How this site is put together
A practical way to pick one agent skill from this directory without installing a pile of prompts you will never run again.
What the install prompt on a skill page actually does, how to tell that it took effect, and how to take it back out.
The cases where a skill card should stay a card: secrets, personal data, licenses you cannot accept, and prompts that fight your agent.
What a DeepSeek Harness plugin is, how it differs from an agent skill, and how to try one without losing the session you were in.
What each DeepSeek Harness plugin category is promising, and the mix-ups that waste an afternoon.
What the fields on a model page mean, which ones are safe to compare, and which ones are just the vendor's sentence.
Editor's picks
Items we have written a full note for, and why they are on the listAn editor's comparison of two flagships at different price points: where the gap is real, where it is a rounding error, and which one belongs in a high-volume pipeline.
The same shortlist keeps appearing in three price bands. A routing decision, not a winner-takes-all comparison, for teams that will end up paying for more than one.
Two listings that both look suited to long inputs, compared on the four fields that actually decide a long-document workload: context, price, license, and hosting.
The same Claude Opus 5 weights are listed at three prices. What you are actually buying with the batch discount and the fast premium, and when neither is worth it.
Three low-price rows that are not competing for the same job. How modality, license, and retry rate decide a high-volume workload that looks like a price comparison.
The complete set of online document operation tools provided by Tencent Docs MCP, supporting creation, querying, and editing of smart docs, Excel, PPT, mind maps, and more.
Bid Proposal Verification Tool is used to verify the consistency between bid proposals and tender documents, checking table of contents, page numbers, attachments, and seals, based on Shibot Technology's bid data API.
A fully local-run WeChat chat analysis tool with MBTI personality inference, sentiment analysis, risk detection, relationship health assessment, and conversation prediction, emphasizing privacy.
An AI skill that generates original lyrics and synthesizes a custom song based on user-provided stories, occasions, and details.
AI agent tool for generating HTML proposals for client visit business scenarios, featuring deep research and dynamic mainline mapping.
Tencent Cloud Table Recognition V3 is an OCR skill for detecting and recognizing tables in images or PDFs, supporting various table types like linear and borderless tables, with Excel export.
Enter a name to auto-research and distill a public figure's cognitive frameworks into an Agent Skill—mental models and heuristics for analysis, not mere role-play.
Install in dsh with no login or API key for unlimited free access to frontier models such as DeepSeek V4.1 Flash and Kimi K3, plus a local OpenAI-compatible gateway.
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
Community DSH addon with 84 host patches, 25-layer self-checks, and a full sidebar panel—local state, snapshots, and rollback to cut confirmation friction for smoother learning and experiments.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
An agent skill where the model writes a Markdown draft and the bundled CLI renders a readable, offline single-file HTML page with diagrams and tables using about one-eighth the output tokens of hand-written HTML.
Claude Opus 5 is Anthropic’s flagship model for demanding reasoning, coding, and long-horizon agentic work. It is particularly strong at end-to-end software tasks, code review and bug finding, visual analysis...
Claude Opus 5 is Anthropic’s flagship model for demanding reasoning, coding, and long-horizon agentic work. It is particularly strong at end-to-end software tasks, code review and bug finding, visual analysis...
Fast-mode variant of [Opus 5](/anthropic/claude-opus-5) - identical capabilities with higher output speed at 2x pricing relative to regular Opus 5. Learn more in Anthropic's docs: https://platform.claude.com/docs/en/build-with-claude/fast-mode
GPT-5.6 Sol Pro is the same underlying model as [GPT-5.6 Sol](https://openrouter.ai/openai/gpt-5.6-sol), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode
GPT-5.6 Sol Pro is the same underlying model as [GPT-5.6 Sol](https://openrouter.ai/openai/gpt-5.6-sol), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode
Grok 4.6 is SpaceXAI's smartest model with frontier performance on coding, knowledge work, and STEM.