OperationsAnswers product questions from your docs, with sources
Docs answerer
Point it at your public docs and the places your team keeps knowledge, and it answers product, feature and how-to questions with a link to the exact page behind each claim. When the docs are silent, thin or out of date it says so plainly instead of guessing, and logs the gap so someone can fix the page. It drafts replies to customers and colleagues for you to send and never sends anything itself.
- Works with
- NotionGoogle DriveGoogle DocsSlackGitHub
- Recommended model
- Claude Opus 5.5
- Skills
- 6
What it does
- Answers product, feature and how-to questions from your live public docs first, with a link on every claim
- Adds what your internal notes, wiki pages and team threads say, labeled apart from the public docs
- Says out loud when the docs do not answer, contradict each other or look out of date
- Drafts replies to customers and colleagues with working links, for you to send
- Keeps a log of doc gaps and conflicts and turns it into a weekly fix list for whoever owns the docs
The skills it brings
Playbooks the mate follows when the job calls for them. You can read and edit every one after adding it.
/dqa-source-mapdqa-source-mapUse when setting up, when the person names a new place where product knowledge lives, or when a source stops loading.
/dqa-answerdqa-answerUse when the person asks any product, feature, command, setup, limit or how-to question about their company's product.
/dqa-draft-replydqa-draft-replyUse when the person pastes a customer email, a ticket or a colleague's message and wants a reply they can send, grounded in the docs.
/dqa-gap-logdqa-gap-logUse when an answer was missing, thin, conflicting or out of date in the docs, or when the person or a routine asks for the doc gaps and a fix list.
/research-source-checkSource checkJudging whether a source can carry a claim: primary or secondary, who published it and why, how old the data is, conflicts of interest, circular citations, triangulating with independent sources, and saying plainly when the evidence is thin.
/content-humanizeRemove signs of AI writingEditing a draft so it stops reading like AI wrote it: find and fix the tell-tale patterns (inflated significance, filler, rule of three, AI vocabulary, em dashes, chatbot phrases), then add a real voice. Use on any draft before it goes out.
Things to ask it
Our docs live at …How do I …A customer is asking: …What are the gaps in our docs this week?