> For the complete documentation index, see [llms.txt](https://docs.questflow.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.questflow.ai/financial-harness/what-is-invest-md.md).

# What Is invest.md?

**Short answer:** `invest.md` is the identity file of a trading agent — the trading equivalent of `CLAUDE.md` or a product `DESIGN.md`. It states strategy, risk tokens, invalidation, and communication style so every session starts from the same judgment, not a blank chat. On Questflow it is the *entry* to the financial harness, not the whole moat.

## Why coding agents got sticky

Codex and Claude Code got useful when they stopped being stateless. A `CLAUDE.md` tells the agent who it is, what it must never do, and how this repo works. Skills, memory, and hooks sit around that file. Without it, every session is amnesia.

Trading needs the same object.

## What belongs in invest.md

A useful `invest.md` is machine-readable enough to enforce and human-readable enough to review:

* **Universe** — what markets the agent is allowed to touch
* **Thesis** — the worldview, not a single ticker
* **Triggers** — what would make it act
* **Sizing** — how much, and what never to exceed
* **Invalidation** — what proves the view wrong
* **Risk tokens** — max leverage, max drawdown, session rules
* **Communication** — how it should explain itself to you or to subscribers

The body is prose. The front matter is tokens the harness can lint, like a linter for risk.

## invest.md is the door, not the house

Any agent can read a markdown file. The harness around it is what compounds:

* `memory/` — decisions, missed trades, regime responses
* `skills/` — composable methodologies
* `plugins/` — data sources
* execution logs and post-mortems

Exporting raw trades is easy. Exporting the agent's intelligence is not the same thing — and that is a product fact, not a threat.

## Day-1 should not be a blank form

A good harness drafts `invest.md` from what it can already see: wallet history, session behavior, implied risk. You confirm, rewrite, or reject. Claimed rules vs actual behavior is useful — that gap is how style-drift detection starts.

## How this shows up in Questflow

Fund managers distill judgment into AI Agents that manage Funds. The public surface is the Fund: thesis, risk, live ledger. Underneath, the harness keeps identity and memory so the agent does not wake up every morning knowing only today's candle.

Related: What is a financial harness? · Coding agent vs finance agent

*Nothing here is financial advice.*