Skills — Complete Guide

Skills are reusable, plain-English instructions — playbooks — that your AllMind assistant follows on its own. Each skill pairs a short when to use this description with a procedure to follow, so the moment a request matches, the assistant applies the right method without you having to spell it out. You start with a large library of expert skills built by AllMind (DCF models, earnings previews, peer comps, sector frameworks, M&A deliverables, and more), and you can teach the assistant your own procedures in your own words — the way you’d brief an analyst once and have them remember it forever. Find it under Research Workspace → Skills.

​

What it is

A skill turns a repeatable analytical method into something the assistant knows and applies consistently. Instead of re-explaining “when I ask for an earnings preview, pull the last four quarters, the consensus setup, the options-implied move, and the key debates” every time, you capture it once and the assistant runs it that way every time. There are two kinds of skills:

• By AllMind skills — a curated library of more than 90 ready-to-use skills, built and maintained by AllMind and grouped by topic (earnings, valuation & modeling, research & diligence, trading & events, sector frameworks, private equity & credit, deal execution & M&A, customer analytics, risk & quant, macro, and product support). These are always available to the assistant — you don’t have to switch any of them on.
• Personal skills — your own playbooks. Create them by describing what you want in a sentence and letting AllMind draft the full procedure, or write the steps yourself. Your personal skills are saved to your account, so they follow you across devices and sessions.

The key idea is automatic application: the assistant reads each skill’s when to use it description and, when your request matches, applies that skill on its own — pulling in the full procedure only when it’s relevant. You can also run any skill on demand when you want to force it. This makes Skills the layer that turns the general assistant into one that already knows your house methods.

​

When to use it

Reach for Skills when:

• You do the same kind of analysis repeatedly — earnings previews, valuation models, peer comps, sector deep-dives, diligence checklists — and want it done the same thorough way every time.
• You want the assistant to handle a specialized task correctly by default — e.g. valuing a bank on capital and returns rather than a generic cash-flow model, or reading a REIT on funds-from- operations instead of earnings per share.
• You have a personal method or house style you want the assistant to remember and apply without being re-briefed.
• You want a finished deliverable produced to a consistent standard — an Excel model, a slide deck, or a Word write-up — built the same way each run.

Use something else when:

• You just want a quick fact or a fast back-and-forth → ask in Chat directly (Skills are how Chat already knows your methods, but you don’t need a skill for a one-off question).
• You want a formatted, templated multi-section report or a presentation-ready document as the main output → Reports (which has its own reusable report templates) or Agent Studio.
• You want the same questions answered across many companies or documents in a table → Grids.
• You have a one-off, open-ended research question that doesn’t recur → just run it in Agent Studio or Chat; a skill pays off when the method repeats.

​

How to use it

​

Using a built-in skill

You usually don’t have to do anything — the By AllMind library is always available, and the assistant applies the right skill automatically when your request fits (“build a DCF for this company,” “give me an earnings preview,” “compare these peers on valuation”). When you want to force a specific skill, there are two ways:

​

Creating your own skill

Click New skill. You have two paths:

​

Managing your skills

From the Personal tab (or the All tab’s “Your skills” section) you can open any of your skills to edit it, delete it (permanent), or toggle Auto-apply on or off. Editing is granular — rename a skill, tweak its procedure, or flip it on or off without touching anything else. Your list is shown newest-updated first.

​

Capabilities & key choices

​

The anatomy of a skill

Every skill has a few parts, each with a clear job:

​

Auto-apply vs. run on demand

• Auto-apply (the default for a new skill). The assistant watches for requests that match the skill’s when to use it description and applies it automatically — you never have to remember it’s there. This is how a skill makes the assistant quietly better at your recurring work.
• Run on demand. Force a specific skill for one request via Use in chat or the / menu. When you run a skill this way it becomes the primary method for that request; the assistant uses your other auto-applied skills only if they directly support it and don’t conflict. Turning a skill’s Auto-apply off keeps it on demand only — useful for a skill you want available but not always firing.

​

Input field types

A starter prompt’s blanks are typed, so you fill each with the right picker: In the editor, inputs are written into the starter prompt with a simple placeholder syntax such as @{ticker:Company} or @{freeText:Question or angle}; AllMind turns each into the matching fill-in control. When AllMind drafts a skill for you, it creates these inputs automatically.

​

The By AllMind library

More than 90 expert skills, grouped by topic. A sampling of what’s there:

​

Product support: the AllMind Support skill

One built-in skill stands apart from the analytical library. AllMind Support doesn’t analyze a company — it answers questions about the AllMind platform itself: how a feature works, how to set it up, troubleshooting, whether something exists, and what shipped recently. It draws on AllMind’s own product documentation and changelog, so the answers stay current as the product evolves. It’s why you can ask, right inside Chat, “how do I build a Grid?”, “how do Data Rooms work?”, “why isn’t my skill firing?”, or “what’s new this week?” — and get an accurate, up-to-date answer without leaving your workflow. Like the other built-in skills it applies automatically when you ask a product question, and you can also run it on demand from the By AllMind library.

​

AI-drafted vs. manual skills

• AI-drafted is the fast path and usually the best: describe the skill in a sentence and AllMind writes a complete, grounded procedure and starter prompt for you to review. Because the draft only references real capabilities and reuses existing skills, it tends to produce a workable skill on the first try.
• Manual gives you full control from the start — best when you already know the exact steps, inputs, and output you want.

Either way you can edit everything afterward, so a good workflow is: draft with AI, then refine the when to use it line and the procedure.

​

What it can access & produce

A skill is a method the assistant follows, so a skill can draw on everything the assistant can reach and produce everything the assistant can build. It can draw on:

• Live market data — prices, fundamentals (income statement, balance sheet, cash flow), ratios, valuation multiples, technicals, and relative performance.
• Estimates & the Street — consensus estimates and revisions, price targets, broker ratings, and analyst coverage.
• Filings & transcripts — company filings (US and Canada), earnings-call transcripts, and sell-side/broker research.
• Events — earnings dates, investor days, and the earnings calendar, plus historical earnings reactions.
• News, sentiment & macro — company and sector news, market sentiment gauges, and official economic indicators.
• Specialized signals — alternative data (consumer card spend, foot traffic, app and web usage, headcount trends), supply-chain and corporate-relationship mapping, prediction-market probabilities, options data, short-interest and institutional-flow signals, and ESG ratings.
• Your own materials — documents in your Data Room, files you upload, and connected sources such as Google Drive — and the web.

It can produce:

• A sourced written answer with the assistant’s reasoning.
• Downloadable files, built to an institutional standard: Excel models and workbooks (with formulas and sensitivity tables), PowerPoint decks in AllMind’s house style, and Word documents.
• Inline charts and tables rendered directly in the conversation.
• A handful of document-assembly skills (for example, a deal “information book”) also assemble a combined PDF of gathered source documents.

​

Tips & best practices

• Nail the “use automatically when” line. It’s what decides when a skill fires. Name the concrete requests, tasks, or kinds of company it covers — and, if it overlaps a built-in skill, say so. A vague description means the skill fires at the wrong times or not at all.
• Let AllMind draft it, then refine. The AI draft grounds the procedure in real capabilities; your job is to tighten the trigger and the steps.
• Keep one skill to one job. A focused “earnings preview” skill beats a sprawling “do all my research” skill — narrow skills fire more reliably and combine better.
• Use the starter prompt to encode your inputs. Put the subject, the decision question, the scope, and the output shape into the starter prompt as fill-in fields, so each run captures what matters.
• Lean on the library. Before building your own, search the By AllMind skills — the method you need may already exist, and you can compose it into your own skill.
• Turn off what you don’t want firing. If a personal skill is too eager, switch Auto-apply off and keep it on demand from the / menu.

​

Limits & things to know

• The description is what triggers a skill. The assistant matches your request against each skill’s when to use it text. If two skills could apply, running one explicitly (Use in chat or /) makes it the primary method for that request.
• Built-in skills are always available to the assistant — there’s no on/off switch for them; the assistant simply applies the relevant one when your request fits.
• Auto-apply is the personal-skill switch. It controls whether your skills fire on their own. Off, they’re still usable on demand.
• Personal skills are private and per-user. They’re saved to your account and visible only to you; they aren’t shared with teammates from here.
• Skills sync with your account. Create a skill on one device and it’s there on the next.
• Local preview mode. If the Skills service is briefly unavailable, you’ll see a “Local preview” notice and your skills save to that browser only. They still work when you explicitly Use in chat (the instructions travel with your message), but they won’t auto-apply or sync across devices until the service reconnects.
• Standard downloadable outputs are Excel, PowerPoint, and Word, plus inline charts and tables. There’s no general PDF or HTML download from a skill (a few document-assembly skills are the exception, assembling a combined PDF).
• Deleting a skill is permanent.
• AI drafting authors the skill, it doesn’t run it. Drafting writes the skill’s name, trigger, procedure, and starter prompt; nothing happens until you save it and use it.

​

How it works with other features

Skills are the method layer that makes the rest of the platform smarter:

• Chat ↔ Skills. This is the home of Skills. The same assistant that powers Chat reads your enabled skills and the By AllMind library and applies the right one as you work — and you can invoke any skill from the chat box with / or from the Skills page with Use in chat.
• Agent Studio ↔ Skills. Agent Studio runs the same assistant autonomously on a task, so it brings your skills and methods to its multi-step research and to the deliverables it builds.
• Reports ↔ Skills. Both make work repeatable, but at different layers: a Skill is a reusable method the assistant applies in conversation; a Report template is a reusable document blueprint that produces a formatted, multi-section report. Use a skill to standardize how analysis is done; use a report template when the deliverable is a structured document. (They are separate features — saving a skill doesn’t create a report template, and vice versa.)
• Grids ↔ Skills. A grid answers the same questions consistently across many companies or documents; a skill standardizes a single analysis or deliverable. Use a skill to define a method, and a grid to apply a consistent question set at scale.
• Data Room & Document Search → Skills. Skills can ground their work in your own materials — gather sources with Document Search, keep them in a Data Room, and a skill’s procedure can read and cite them alongside market data.

Inputs that feed a skill: the starter prompt’s fields (company, period, scope, your question), your own documents and connected data, and everything the assistant can access. Outputs a skill hands off: a sourced answer and downloadable Excel/PowerPoint/Word files — ready to share or to feed the next step.

​

Example workflows

​

1. Teach the assistant your earnings routine once, reuse it every quarter

Goal: a consistent earnings preview on any name, your way.

1. New skill → describe it: “An earnings preview that pulls the last four quarters of estimates vs. actuals, summarizes the setup into the print and the key debates, flags consensus revisions in the last 30 days, and ends with the bull and bear scenarios and what to watch on the call.”
2. Review the draft — confirm the use automatically when line (“when I ask for an earnings preview on a stock”) and the starter prompt’s inputs (company, period).
3. Save with Auto-apply on. From now on, whenever you ask for an earnings preview the assistant runs it your way — no re-briefing.
4. When you want it explicitly, open the skill and Use in chat, fill in the company, and send.

​

2. Analyze a sector and build the deliverables — with the right frameworks applied automatically

Goal: analyze the gold sector and produce a model and a deck.

1. Gather sources — use Document Search to find the relevant filings, transcripts, and research, and collect the best into a Data Room.
2. Compare the producers — build a Grid to answer the same questions (margins, leverage, cost position, key risks) across the names consistently.
3. Model the leaders — in Chat, ask for a valuation on the top one or two names. The DCF and relevant sector skills apply automatically (mining frameworks, not a generic model), and the assistant returns a downloadable Excel model.
4. Go deep where it matters — run Agent Studio on the one or two themes that need a full narrative; your skills travel into that work too.
5. Package it — ask for a summary deck; the assistant produces a PowerPoint in house style. For a formatted written report, use Reports with a sector or peer-comparison template.

The result: a consistent, sourced analysis plus finished deliverables — with the right methods applied at each step because the skills are doing the remembering.

​

3. Standardize a deal deliverable across the team’s workflow

Goal: a repeatable diligence request list for every new deal.

1. New skill → describe it: “Generate a workstream-by-workstream due-diligence request list for a target, tailored to the deal type and sector.”
2. Review and save. The starter prompt asks for the target and the deal type.
3. Run it on each new deal via Use in chat — you get the same structured request list every time, as a downloadable document, instead of rebuilding it from scratch.

​

Common questions

​

Getting help

For help building a skill or designing a workflow around one, reach AllMind support through the in-app support option or your account team.