Daily picks · November 17, 2026
Bots With a Job to Do
This week's picks share a common thread: they do not just chat, they take a task from handoff to a verified finish, whether that means warming up a persona, tidying a vault, or reviewing a draft.
A Warm Butler With Opinions
大肥鱼 is a Tobs persona export by wanghenry9 that doubles as a complete butler assistant. It is gentle without slipping into customer-service tone: it gives one recommendation instead of a menu of options, keeps replies to one or three sentences, and holds three hard lines of never fabricating, never bluffing, and never performing. The listing ships a full persona file, a Core Memory template, and appearance assets, so it works both as a ready assistant and as a reusable sample for building your own. It is the most-installed bot of the week.
A Steward for Your Obsidian Vault
Built on the LLM-Wiki approach, this bot imports sources while preserving the originals, summarizes them, and links related topics, concepts, and entities. It answers questions from the knowledge base with citations and disclosed uncertainty, and runs reviews for broken links, orphan pages, duplicates, missing sources, conflicts, and outdated content. It backs up before making changes and never silently overwrites conflicting sources. A strong pick for anyone whose vault has outgrown manual upkeep.
The Manager Who Actually Delivers
言言 positions itself as a team general manager: it restates the desired deliverable in one sentence, asks all clarifying questions at once, then breaks work into single-responsibility subtasks with checkable completion criteria. It routes retrieval to cheap assistants and code changes to tool-equipped executors, verifies each result against evidence such as command output or file paths, and reports back in a fixed four-line format. The persona is a full workflow playbook, which makes it stand out for multi-step delivery.
A Reviewer That Leaves the Pen Down
A rising bot that reviews written materials such as curriculum standards, reports, plans, and papers across four dimensions: authorial voice, argument direction, material completeness, and formatting conventions. It issues a three-tier verdict of acceptable, needs revision, or needs rewrite, with executable Major comments that cite section numbers and original sentences, plus numbered Minor suggestions. It only gives opinions and never rewrites, and it flags uncertain facts for verification. A careful second pair of eyes for formal documents.