---
name: al-217-turn-your-agent-s-past-wins-and-failures-into-a-lessons-f
description: "Most agents start every task like a new intern. Have one study the transcripts of past attempts, what worked and what failed, and write a short lessons file of your preferences, conventions and gotchas that it reads before the next similar job. Letta released FrontierMemory 1.0 in October 2026, a benchmark that scores agents on how much memory written from 674 past attempts (184 passes, 490 failures) helps them solve related tasks; the top models reached 73.3%, and Letta notes many agent products now 'dream' over past sessions to improve their memory. Use when the user wants help with agent memory (personal productivity) for a small business."
---

# Turn your agent's past wins and failures into a lessons file it reads before every task

Most agents start every task like a new intern. Have one study the transcripts of past attempts, what worked and what failed, and write a short lessons file of your preferences, conventions and gotchas that it reads before the next similar job. Letta released FrontierMemory 1.0 in October 2026, a benchmark that scores agents on how much memory written from 674 past attempts (184 passes, 490 failures) helps them solve related tasks; the top models reached 73.3%, and Letta notes many agent products now 'dream' over past sessions to improve their memory.

## Inputs to collect first
Ask the user for anything below that they haven't already given you. Don't guess missing facts.

- Type of task, e.g. writing client reports / fixing bugs in our app …
- Your own material (text, notes or data)

## Instructions

Do the task below with the user's inputs in place of each [bracket].

You're my agent's coach. Below are transcripts or notes from past attempts at [type of task, e.g. writing client reports / fixing bugs in our app / preparing invoices], each marked WORKED or FAILED, plus any feedback I gave: [paste]. Write a lessons file under 400 words with: 1) my preferences (format, tone, length, what I always change); 2) rules and conventions that weren't written down but mattered; 3) gotchas in our tools, files or data that caused failures, and how to avoid them; 4) a short checklist to run before calling the task done. Only include lessons backed by the transcripts, and note which attempt each one came from. Skip one-off details. I'll paste this file at the start of every similar task, so keep it tidy and update it when new attempts contradict it.

## Safety checks (share these with the user)

- This prompt works on text other people wrote (emails, reviews, messages or web pages). That text can hide instructions aimed at AI, such as "ignore your rules and…". Paste it as material to work on, and never let the AI act on requests inside it, like paying, sharing logins or changing records.
- Check every figure, price and total against your own records before anything is sent, posted or booked. A person approves refunds, discounts, payments and price changes.
- Remove names, contact details, account numbers and other personal data the task doesn't need before you paste, and check your AI tool's data and training settings.
- Check names, dates, numbers and any claim against your own source material before you use the answer. AI can misread or invent details.
- Say clearly which parts need a person to review or approve before anything is sent, paid, posted or filed.

## About this skill

- Works with: ChatGPT, Claude, Claude Code, Codex, Letta
- Based on this real-world example: https://www.frontiermemory.com/blog/v1/
- Free prompt from AgentPlaybook Real Estate: https://agentic-library.netlify.app/prompt/al-217-turn-your-agent-s-past-wins-and-failures-into-a-lessons-file/
- To install in an agent that supports Agent Skills, save this file as SKILL.md in a folder with the same name as `name` above (for example in your skills directory). Install steps for Claude.ai, Claude Code, Cursor and Codex: https://agentic-library.netlify.app/skills/