reflection-coach

Analyze agent execution records to propose evidence-backed instructions and skill improvements for performance optimization.

79.8k|14.6k|Updated Mar 2, 2026
One-click install
npx skills add https://github.com/paperclipai/paperclip --skill reflection-coach
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: reflection-coach
Source: https://github.com/paperclipai/paperclip/tree/main/packages/skills-catalog/catalog/bundled/paperclip-operations/reflection-coach
Command: npx skills add https://github.com/paperclipai/paperclip --skill reflection-coach

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill addresses the challenge of agent performance degradation by identifying patterns in past execution failures and proposing targeted, evidence-based improvements to agent instructions and tool usage.

Core Features & Use Cases

  • Evidence-Based Coaching: Analyzes issue trajectories and comments to identify specific failure clusters like verifier-misses or instruction-misses.
  • Durable Improvement: Proposes minimal, high-impact changes to AGENTS.md or reusable skills to prevent recurring issues.
  • Safe Workflow: Implements a strict gatekeeping process requiring reviewed diffs and accepted task interactions before any changes are applied.

Quick Start

Use the reflection-coach skill to analyze the last 20 issues for agent 'coder-bot-01' and propose improvements for identified failure patterns.

Frequently Asked Questions about reflection-coach

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I coach an agent using this skill?

Provide the targetAgentId and a window of issues to the reflection-coach. The skill will analyze the agent's recent execution record, cluster failure patterns, and generate a proposal document with specific diffs for the agent's instructions or skills.

Can I apply changes automatically with this skill?

No, this skill enforces a strict separation between discovery and application. You must review the generated diff and obtain an accepted task interaction before any changes can be applied to the agent's configuration.

What happens if the agent already has the proposed rule?

The skill performs a pre-analysis of the agent's current AGENTS.md and assigned skills. If a rule is already present, it identifies the failure as a compliance issue rather than a missing instruction, proposing ways to make the existing rule more effective.

When should I avoid using this skill?

Do not use this skill to reflect on yourself, to rewrite shared infrastructure code, or to apply changes without a reviewed diff and an accepted interaction. It is strictly for coaching other agents based on evidence-backed trajectories.