Automation Mining
An open-source Agent Skill released by n8n Labs that mines hidden automation opportunities from a team's day-to-day activity data. Built on a taxonomy and a large dataset of workflows, it ranks opportunities by frequency multiplied by time cost, helping AI coding agents identify high-value repetitive tasks.
Timeline (last 90 days)
n8n Labs has open-sourced an Agent Skill called Automation Mining.
Automation Mining is essentially an instruction file in Markdown format (an Agent Skill) that AI coding agents such as Claude Code, Codex, and Cursor can read and execute directly.
The Automation Mining workflow has four steps: collect data, classify and match, score and rank, and propose solutions.
Automation Mining uses "frequency × time cost" as its scoring and ranking function, converting repetitive tasks into quantifiable monetary losses.
n8n Labs tested with 57 synthetic companies, comparing a baseline agent with and without the Skill, and task recall rose from 28.2% to 51.7%.
In the same test, proposal quality rose from 44.7% to 53%, an increase of 8.3 percentage points.
The Automation Mining Skill uses a read-only design: it does not publish content, does not edit anything, and does not write to the system; it only observes, analyzes, and proposes.
The Automation Mining Skill uses the Apache 2.0 open-source license, and the repository address is n8n-io-automation-mining-skill.
Automation Mining has four limitations: synthetic data underestimates real-world complexity, the recall ceiling is about 51.7%, the read-only design means users must build the workflows themselves, and the product is too new and may have rough edges and missing documentation.
All Facts (9)
Automation Mining has four limitations: synthetic data underestimates real-world complexity, the recall ceiling is about 51.7%, the read-only design means users must build the workflows themselves, and the product is too new and may have rough edges and missing documentation.
50%UnverifiedAutomation Mining is essentially an instruction file in Markdown format (an Agent Skill) that AI coding agents such as Claude Code, Codex, and Cursor can read and execute directly.
50%UnverifiedThe Automation Mining workflow has four steps: collect data, classify and match, score and rank, and propose solutions.
50%UnverifiedAutomation Mining uses "frequency × time cost" as its scoring and ranking function, converting repetitive tasks into quantifiable monetary losses.
50%Unverifiedn8n Labs tested with 57 synthetic companies, comparing a baseline agent with and without the Skill, and task recall rose from 28.2% to 51.7%.
50%UnverifiedIn the same test, proposal quality rose from 44.7% to 53%, an increase of 8.3 percentage points.
50%UnverifiedThe Automation Mining Skill uses a read-only design: it does not publish content, does not edit anything, and does not write to the system; it only observes, analyzes, and proposes.
50%UnverifiedThe Automation Mining Skill uses the Apache 2.0 open-source license, and the repository address is n8n-io-automation-mining-skill.
50%Unverifiedn8n Labs has open-sourced an Agent Skill called Automation Mining.
50%