Can I manually trigger a review of learning opportunities?

Yes, you can use the /coach review command to see pending proposals or /coach scan to proactively check for environment-wide improvements.

Where are the learned rules stored?

Approved rules and guidelines are typically added to the CLAUDE.md file in your project or global configuration directory, making them persistent across sessions.

Does Claude Coach update my files automatically?

No, Coach adheres to a 'no silent writes' principle; all proposed improvements must be reviewed and approved by the user via the /coach approve command.

What types of signals does Coach detect?

It monitors for command failures, explicit user corrections, tone escalation (frustration), repeated instructions, and outdated tool warnings.

Claude Coach

Name: Claude Coach
Author: netresearch

bynetresearch

생산성 및 워크플로

Empowers Claude to learn from user corrections and tool failures by automatically proposing rule updates and workflow improvements.

소개

Claude Coach is an intelligent learning layer that monitors your interactions to identify friction points, such as repeated instructions or command errors. Instead of making the same mistake twice, Coach extracts actionable improvement candidates—like new rules, checklists, or code snippets—and presents them for your approval via the /coach command. By systematically updating your project's CLAUDE.md file, it ensures that Claude's behavior aligns perfectly with your specific preferences and project requirements over time.

주요 기능

Multi-Level Scoping: Intelligently determines if improvements should apply globally or to specific project contexts.
Automatic Friction Detection: Identifies user corrections, tone escalation, and repeated instructions in real-time.
Human-in-the-Loop Approval: Ensures no silent changes by requiring explicit user consent via /coach approve.
Error-Driven Learning: Converts command failures and stderr patterns into actionable project rule updates.
Proactive Tool Scanning: Detects outdated CLI tools and dependencies to prevent version-related issues before they occur.
0 GitHub stars

사용 사례

Preventing Claude from repeating mistakes after you've corrected its coding style or file structure choices.
Keeping development environments up-to-date by identifying deprecated tools and libraries during routine scans.
Automatically generating checklists for complex deployment or testing workflows based on previous session data.

소개

주요 기능

Multi-Level Scoping: Intelligently determines if improvements should apply globally or to specific project contexts.
Automatic Friction Detection: Identifies user corrections, tone escalation, and repeated instructions in real-time.
Human-in-the-Loop Approval: Ensures no silent changes by requiring explicit user consent via /coach approve.
Error-Driven Learning: Converts command failures and stderr patterns into actionable project rule updates.
Proactive Tool Scanning: Detects outdated CLI tools and dependencies to prevent version-related issues before they occur.
0 GitHub stars

사용 사례

Preventing Claude from repeating mistakes after you've corrected its coding style or file structure choices.
Keeping development environments up-to-date by identifying deprecated tools and libraries during routine scans.
Automatically generating checklists for complex deployment or testing workflows based on previous session data.