Skills did not start in investing. They came out of the world of AI coding agents, where people kept hitting the same wall: a model is brilliant in the moment and forgetful by design. It can write a function beautifully, then forget your conventions on the very next one. The fix that emerged was to write the recurring knowledge down once, in a form the agent could pick up on its own.
Anthropic gave that pattern a name and a shape. A skill is a small folder with a SKILL.md file inside: a name, a description of when to use it, and the steps to follow. They released it as an open standard, now documented at agentskills.io and supported across a growing list of agents. It spread quickly for a simple reason. It let people stop repeating themselves and start compounding what they knew.
Investing has exactly the same shape, and almost nobody treats it that way. A good investor is not just smart in the moment. They run a process: the checks they always do before trusting a number, the order they read a business in, the questions they ask before sizing a position. That process is the edge, and it is the part a raw model does not have. It has the reasoning. It does not have your method.
Out of the box the Investi analyst is deliberately neutral. It answers however you ask, which is useful, but it means your process lives outside the tool, in your head, retyped into every chat.
Skills are how you hand the analyst your method, so it works the way you do, every time.
Skills are live in the beta today. Here is what they are, and why they matter even more in investing than in the code editors they came from.
A skill is a workflow, not a prompt
A prompt is a single instruction. You type it once, you get one answer, and it is gone. A skill is that same knowledge made durable: written down once, then picked up on its own whenever it fits. Once it exists, the analyst can reach for it without being told.
Say you write an earnings review skill. From then on, when you ask about a company that just reported, the analyst notices the skill fits the moment and runs your checklist: revenue against guidance, the margin trend, what changed in the language on the call, the two or three numbers you actually care about. You did not paste anything. It recognized the situation and applied your method.
“How did NVDA do this quarter?”dcf-valuationearnings-deep-divemoat-analysisdividend-safetymatched by its descriptionRuns your earnings checklist, step by stepEnabled skills are matched to the momentThere are two ways a skill runs. It gets picked up automatically when it fits what you are asking, which is the whole point. Or you can trigger any skill by hand by typing / in chat and choosing it, even one you keep switched off for everyday use.
How did NVDA do this quarter?Applied skillearnings-deep-diveThoughtRead TranscriptGet FinancialsRevenue beat, but guidance is the real story.Data center revenue landed at $30.8B against a $28.7B consensus, and gross margin held at 75%. Management guided next quarter above the Street.Per your checklist: no one time items, tone stayed confident, and the only guidance change was an upward revision to full year capex.The analyst applies your skill on its ownWhat is inside a skill
Under the hood, a skill is a small text file called SKILL.md. It holds three things that matter: a name, a description of when to use it, and the body, which is the actual workflow written in plain language. You can also attach reference files, like a checklist, a scoring rubric, or a note template, that the analyst pulls in when the skill needs them.
The description carries more weight than it looks. It is how the analyst decides a skill applies, so the best ones read like an instruction to a colleague: use this when a company reports, use this before I size a position, use this when I am comparing two names in the same sector.
SKILL.mdnameearnings-deep-divewhat to call itwhen to usea company just reportedhow the analyst finds itstepsread, compare, flag, summarizewhat it actually doesoptionalchecklist.mdtemplate.mdpulled in when neededA skill is a small text file, SKILL.md SKILL.mdchecklist.md1---2name: earnings-deep-dive3description: Use this when a company reports.4---5 6# Earnings deep dive7 81. Read the transcript and the quarterly filing.92. Flag guidance changes and one time items.103. Note any shift in management tone.114. Compare the print to consensus, then summarize.Author a skill in plain languageGetting one in seconds
You do not have to start from a blank page. There are three ways to add a skill, and all of them take about a minute.
Write your own directly in the editor. Import one from a GitHub repository by pasting a link. Or upload a file, either a .zip of one or more skills or a single SKILL.md. Community skills mean you can borrow a solid workflow and shape it to fit how you actually think.
Because Investi uses that same open format, skills are portable in both directions. Any skill written for the standard drops straight into Investi, so you inherit a whole ecosystem instead of a walled garden. And the skills you write here are yours to take anywhere, because they are plain text files, not something locked to us.
Write your ownin the editorImport from GitHubpaste a repo linkUpload a file.zip or SKILL.mdYour skill libraryStart from the community or a blank pageEvery skill has a switch. The ones you leave on are the ones the analyst discovers on its own. Anything you switch off stays out of its way but is still there when you want to run it by hand. So your library can be broad without making every answer noisy.
SkillsSearch or paste a link…dcf-valuationValue a company with a full DCF: FCF history, a five year forecast, and a stress tested terminal multiple.earnings-deep-diveRead the latest transcript and quarterly filing, then flag guidance changes, one time items, and shifts in management tone.moat-analysisScore durable advantages across switching costs, network effects, scale, and brand before trusting a narrative.bear-case-red-teamArgue the short thesis: surface the three most likely ways the position loses money.Your skill library, on or offSkills steer the work, your judgment stays yours
A skill decides how the analyst works. It does not decide what it is allowed to change on your behalf. That line matters, and we do not cross it.
When a skill leads the analyst to touch your notes, say it wants to fold a new quarter into your thesis, the change still shows up as a diff you approve or reject. Here is what your notes say, here is what would change, nothing is written until you say so. Skills make the analyst more opinionated about method while leaving you in full control of the record.
AAPL / Investment Thesis+3−1## Key numbers (FY24)## Key numbers (Q1 FY26)- Revenue: $124.3B (+4% y/y), EPS: $2.40- Services: $26.3B (+14% y/y), all-time highRejectApproveEvery change still waits for your approvalWhy this matters
We wrote before that the durable work in AI is not the model, it is the layer around it: the context, the tools, the memory, the trust. Skills are the first piece of that layer you get to write yourself.
The model brings the reasoning. It can read a filing and follow a call as well as anyone. What it does not have is your taste, the accumulated sense of what to look at first, what to distrust, and what to ignore. That lived in your head, and in your head it quietly degraded. A skill is where you put it down so it stops slipping away, and so the analyst can carry it for you.
The model brings the reasoning. The skill brings your rigor. Together they run your process, every time.
Skills are live in the beta now. Browse the community library or write your first one on the features page, and read Anatomy of a harness for why we think this layer, not the model, is the real product.