What is the RLM pattern used by Atris Memory?

RLM stands for 'grep first, reason second.' It is a search strategy that uses fast, local grep commands to find matches before using LLM reasoning, which optimizes for speed and reduces token costs.

How does Atris store project history?

History is stored in a structured markdown format within the 'atris/logs/' directory, organized by year and date to allow for easy chronological retrieval.

When should I use the Atris Memory skill?

Use this skill whenever you need to recall why a decision was made, what features were previously tried, or when you need to identify patterns in project failures or technical debt.

How does the skill handle a large volume of historical data?

If a search returns more than 10 matches, the skill automatically escalates to using specialized subagents to chunk, analyze, and synthesize the data into a concise summary.

Atris Context Memory

Name: Atris Context Memory
Author: atrislabs

byatrislabs

•

生产力与工作流

Manages and retrieves historical project context from Atris journals to help agents reason over past decisions and work patterns.

Atris Context Memory provides Claude Code with a persistent, searchable history of past work, architectural decisions, and project evolution. By implementing the 'grep first, reason second' (RLM) pattern, it allows agents to efficiently scan daily journals for specific keywords before escalating to semantic analysis via subagents. This skill is essential for maintaining project continuity, understanding why specific technical choices were made, and identifying recurring failure patterns or technical debt across the entire development lifecycle.

主要功能

01RLM Pattern search strategy for high-performance, cost-efficient historical retrieval

02Automated pattern recognition for recurring technical debt and project failures

03Pre-configured grep patterns for tracking decisions, ideas, and technical pivots

04Deep integration with structured Atris journals including Inbox, In Progress, and Completed states

05Multi-layered analysis using Haiku subagents for complex historical synthesis

0654 GitHub stars

使用场景

01Investigating the historical rationale behind specific architectural or code decisions

02Synthesizing progress reports by aggregating completed tasks across multiple journal entries

03Analyzing recurring failure reasons in reviews to improve project quality

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add atrislabs/atris memory

For use in Claude.ai and ChatGPT

Download Skill