Paper 2603.18743
Memento-Skills: Let Agents Design Agents
- Published
- Mar 2026
- Research lab
- Independent
- Citations
- 39
- GitHub
- 1.5K stars
01 In brief
Summary
Memento-Skills is a generalist LLM agent system that functions as an agent-designing agent, autonomously constructing, adapting, and improving task-specific agents through experience.
It uses a memory-based reinforcement learning framework with stateful prompts, where reusable skills stored as structured markdown files serve as persistent, evolving memory.
The system employs a Read–Write Reflective Learning mechanism: in the read phase, a behaviour-trainable skill router selects the most relevant skill; in the write phase, the agent updates and expands its skill library based on new experience.
This closed-loop design enables continual learning without updating LLM parameters.
Experiments on the General AI Assistants (GAIA) benchmark and Humanity's Last Exam (HLE) demonstrate sustained gains, achieving 26.2% and 116.2% relative improvements in overall accuracy, respectively.
The system outperforms a static Read-Write baseline by 13.7 percentage points on GAIA and 20.8 points on HLE.
The skill library grows from 5 atomic skills to 41 skills after GAIA learning and 235 skills after HLE learning, clustering into semantically meaningful domains.
The router, trained via single-step offline RL with InfoNCE loss, improves Recall@1 from 0.54 (Qwen3) to 0.60 and judge success rate from 0.79 to 0.80.
02 From the paper
Abstract
We introduce \emph{Memento-Skills}, a generalist, continually-learnable LLM agent system that functions as an \emph{agent-designing agent}: it autonomously constructs, adapts, and improves task-specific agents through experience. The system is built on a memory-based reinforcement learning framework with \emph{stateful prompts}, where reusable skills (stored as structured markdown files) serve as persistent, evolving memory. These skills encode both behaviour and context, enabling the agent to carry forward knowledge across interactions. Starting from simple elementary skills (like Web search and terminal operations), the agent continually improves via the \emph{Read--Write Reflective Learning} mechanism introduced in \emph{Memento~2}~\cite{wang2025memento2}. In the \emph{read} phase, a behaviour-trainable skill router selects the most relevant skill conditioned on the current stateful prompt; in the \emph{write} phase, the agent updates and expands its skill library based on new experience. This closed-loop design enables \emph{continual learning without updating LLM parameters}, as all adaptation is realised through the evolution of externalised skills and prompts. Unlike prior approaches that rely on human-designed agents, Memento-Skills enables a generalist agent to \emph{design agents end-to-end} for new tasks. Through iterative skill generation and refinement, the system progressively improves its own capabilities. Experiments on the \emph{General AI Assistants} benchmark and \emph{Humanity's Last Exam} demonstrate sustained gains, achieving 26.2\% and 116.2\% relative improvements in overall accuracy, respectively. Code is available at https://github.com/Memento-Teams/Memento-Skills.