Skills

Out of the box, the analyst is unopinionated. It answers however you ask. A skill is a reusable process you give it: the checks you want run, the shape of the write-up, the standards you do not want to retype. A skill might encode value-investing checks, the way your firm writes an initiation, or a due-diligence checklist entirely your own. The philosophy is always opt-in: chosen, visible, and removable. Never a default. A skill steers how the analyst works. It does not silently change your workspace. Edits to notes and skills still go through approval.

How skills run

Automatically

When it’s relevant

Enabled skills are discovered by the analyst when they match what you asked.

Manually

On this message

Type / and the skill’s name to run any skill, even one you’ve disabled for automatic use.

Three ways to get a skill

Catalog

Install one

Pick a catalog skill and run it on a company.

Chat

Ask the analyst

Describe the process. Review the draft like a note.

Editor

Write it yourself

Open the editor and write the instructions.

You can also import a public repository or upload a file. Those are the same skill, arriving a different way.

Ask the analyst to write one

You do not have to open the editor. Describe the process in chat the way you would describe it to a colleague. The analyst drafts the skill and you review the proposal the same way you review a note.

Write a skill I can reuse on every earnings print. Pull the quarter against the same quarter last year and against consensus, read the transcript for guidance changes, and end with whether the thesis moved. Call it earnings-print.

You can also do this after a run you liked:

Turn the checks you just did into a skill I can reuse on the next print.

Open the draft, read the instructions, and approve it only if you would actually want those steps followed. A skill is instructions your analyst will follow.

The Skills view

Open Skills in the left navigation to manage everything:

A single search box filters your installed skills and the catalog, and accepts a pasted GitHub URL for importing.
Each skill has an enable/disable toggle (for automatic discovery) and opens in the editor on click.
Browse skills shows the curated catalog, with one-click installs like Thesis tracker, Financial statement analysis, Earnings analysis, Initiating coverage, Comparable companies, and Due-diligence checklist. Catalog skills are plain instructions like any other, so you can open one in the editor after installing and read exactly what it tells the analyst.

Installing skills

From the catalog

Select Browse skills to open the curated catalog. Each skill installs with one click. Try it: install Earnings analysis, then in chat:

latest quarter

From GitHub

Paste a GitHub repository URL into the search box. Investi checks the repository (public repositories only), then shows a preview of every skill it discovered, each with a checkbox and any warnings. Confirm with Import N skills.

Private repositories can’t be imported; the check will report “Repository not found” for them.

By upload

Use the upload button to import a .zip archive or a single SKILL.md file, up to 10 MB. A zip can contain a full skill with its reference files and assets.

Where to find skills

Skills are plain-text files, so a public repository of them is importable. Places people look:

Maintained set

anthropics/skills

A small set from the team that published the common SKILL.md format.

After installing

In the Skills view	What it does
Toggle on	The analyst discovers it automatically when relevant
Leave off	Manual only. You can still run it with `/` in chat
Select it	Opens the editor so you can read or adjust the instructions
Trash icon	Deletes the skill

Writing your own skill

Create a skill

In the Skills view, select + to open the skill editor with a starting template:

SKILL.md

---
name: my-skill
description: What the skill does and when the agent should use it.
---

The full instructions the agent follows when this skill is active.

The editor shows SKILL.md plus any bundled reference files, with markdown editing, preview, and the same diff review used for notes. A skill is a plain-text SKILL.md file: a short header with a name and description, then the instructions the analyst follows. You can bundle extra files next to it. Because the format is an open standard, you can bring a skill in from elsewhere or take yours out.

Frontmatter rules

The block at the top of the file, between the --- lines, is how the analyst finds and names the skill.

Field	Rules
`name`	Required. Lowercase letters, numbers, and hyphens; max 64 characters; unique across your skills. It becomes the `/` slug in chat.
`description`	Recommended (warned if missing); max 1024 characters. This is how the analyst decides when the skill is relevant. Make it say when to use it, not just what it is.

Reference files and assets

Skills can bundle extra files next to SKILL.md: a methodology document, a report template, a checklist. The analyst can read them when running the skill, and you can hand one directly to the analyst in any chat with mentions. Binary assets (images, PDFs, spreadsheets) are kept read-only with preview and download.

What a good skill looks like

A good skill reads like a process you would hand a colleague, not like a product description.

Write	So that
Describe when, not just what	Automatic discovery has something to match. “Use when analyzing a new earnings report” beats “Earnings analysis”
Numbered steps with explicit outputs	Runs stay consistent: “produce a table of…”, “end with a verdict…”
Thresholds, red flags, required checks	The analyst applies your standards every time

SKILL.md

---
name: earnings-print
description: Use when a company has just reported, or when the user asks for an earnings recap or print review.
---

1. Pull revenue, EPS, and gross margin versus the same quarter last year and versus consensus.
2. Read the transcript for guidance changes and any new risk the company named.
3. Compare those points to the company's thesis note if one exists.
4. End with a short verdict: thesis supported, weakened, or unchanged, and why.

Skills, MCP, subagents, and monitors

Skill

A process you give the analyst

The same checks every time

MCP

A connected source of tools or data

A source Investi does not ship

Subagent

A built-in specialist

A noisy job. You do not write these.

Monitor

A chat request on a schedule

The question asked again without you there

A skill can still cause a delegation: on is the parent’s process. Pulling several years of statements can still go to the Analyst specialist.

Getting started

The AI analyst

Companies

Organizing your research

Customize

Account

How skills run

Three ways to get a skill

Ask the analyst to write one

The Skills view

Installing skills

From the catalog

From GitHub

By upload

Where to find skills

After installing

Writing your own skill

Create a skill

Frontmatter rules

Reference files and assets

What a good skill looks like

Skills, MCP, subagents, and monitors

​How skills run

​Three ways to get a skill

​Ask the analyst to write one

​The Skills view

​Installing skills

​From the catalog

​From GitHub

​By upload

​Where to find skills

​After installing

​Writing your own skill

​Create a skill

​Frontmatter rules

​Reference files and assets

​What a good skill looks like

​Skills, MCP, subagents, and monitors

How skills run

Three ways to get a skill

Ask the analyst to write one

The Skills view

Installing skills

From the catalog

From GitHub

By upload

Where to find skills

After installing

Writing your own skill

Create a skill

Frontmatter rules

Reference files and assets

What a good skill looks like

Skills, MCP, subagents, and monitors