Hedge Layer/DocsOpen app

Hedge Layer Agent memories and skills

Manage these in the Hedge Layer Agent page. Hedge Layer Agent reads them at the start of every turn in the desktop app, on WhatsApp and in scheduled runs.

Memories

Short notes, up to 300 characters each and 50 in total, that Hedge Layer Agent keeps between conversations: the venues and markets you follow, how you size positions, the bankroll to assume for Kelly calculations, or how you like answers written. When you ask Hedge Layer Agent to remember something, or mention a lasting preference, Hedge Layer Agent shows the suggested note and saves it only after you choose Remember. You can also add and delete memories in settings.

Hedge Layer Agent never suggests remembering prices, probabilities, news or anything it read from data sources. Those change, so it fetches them fresh instead.

Research log

The latest probability estimate Hedge Layer Agent calculates for each market in a completed turn is saved to your research log: the market, its price at the time, the estimate and range, the verdict, and the starting probability, factor explanations, and sources behind it. When you come back to a market or topic, or ask what Hedge Layer Agent concluded before, it looks up the earlier estimate and compares it with fresh data, telling you when it was made and what has changed. A saved estimate is never reused as a current number.

Your saved estimates remain until you delete them. The Research section of the Hedge Layer Agent page shows the full history, including the reasoning and sources. Hedge Layer Agent automatically recalls up to the newest 100 records that fit its context; older entries remain in that section. You can delete one or clear them all.

Skills

A skill is a saved workflow in the open Agent Skills format, the same SKILL.md format used by Claude, Codex, Copilot and other agents. Hedge Layer Agent sees each enabled skill's name and description. When your request matches one, or you name it ("run my weekly-fed-check"), it reads the full instructions and follows them. Skills pair well with schedules: a schedule prompt can simply say which skill to run.

Write a skill in settings, or paste a SKILL.md:

---
name: weekly-fed-check
description: Weekly review of Fed-rate prediction markets and what moved them
---

1. Search Polymarket and Hyperliquid for markets on the next FOMC decision.
2. Show implied probabilities for each outcome in one table, with the change since last week.
3. Search evidence for the two biggest moves and give one line on each.
4. Finish with what to watch before the meeting.

Names use lowercase letters, digits and hyphens. Instructions can be up to 8,000 characters, with up to 20 skills per account. Hedge Layer uses only the name, description and instructions; bundled scripts and reference files are not supported.

Limits on what they can change

  • Your memories and skills set preferences. They cannot turn on disabled data sources, add tools, skip the step where you choose markets to research, or make Hedge Layer Agent trade.
  • Hedge Layer Agent follows skill instructions, so read a skill from someone else before you add it.
  • Only a signed-in browser can view or change your skills and research log. Apps that use your token can list memories and save the ones you confirm.
  • Everything is deleted with your account, together with your Hedge Layer Agent conversation history.