AIFlare
VisionDocs
Sign inStart free
Getting Started
  • Introduction
  • Quickstart
Features
  • Project Dashboard
  • Project Settings
  • Timeline
  • Team Collaboration
Skills
  • Skills Overview
  • Context Capture
  • Code Intent Lookup
  • Context Inject
  • Daily Digest
  • Weekly Digest
  • Session Compare
  • Session Summary
  • Prompt Evaluation
Administration
  • Members & Invitations
  • Account Settings
Insights

Prompt Evaluation

Evaluates the quality of prompts used in a session against best practices. Find out what got stuck and how to rewrite it — with Before → After examples.

Overview

Prompt Evaluation analyzes the prompts sent by the user to the AI in a Work Session and rates their quality. It gives concrete feedback on which instructions worked well and which caused confusion, helping you improve how you communicate with AI agents.

Report Structure

Each generated report includes the following sections.

01
Overall Quality Score

Clarity, specificity, and context-richness scores for the session's prompts

02
Effective Prompts

Patterns from instructions the AI understood quickly and executed well

03
Prompts Needing Improvement

Segments that required repeated correction or caused confusion, with root cause analysis

04
Before → After Rewrite

Revised versions of underperforming prompts written in a more effective style

Usage Guide

Command Syntax

Type the command below in the Claude Code conversation window. Omitting sessionId uses the current session automatically.

/prompt-evaluate [sessionId?]

Parameters

NameTypeRequiredDefaultDescription
sessionIdstringOptionalcurrent sessionID of the session to evaluate. Omit to use the current Claude Code session (CLAUDE_SESSION_ID env var). If absent, the skill also searches for .claude-prompts-* files automatically.

Execution Flow

1The skill prompt loads and resolves sessionId: explicit argument → CLAUDE_SESSION_ID env var → .claude-prompts-* file search.
2If no session data is found, the skill outputs 'no captured data' and exits early.
3The get_session_prompts tool fetches the prompt-response pairs from the session. The LLM generates a Markdown report with overall quality score, effective prompts, prompts needing improvement, and Before → After rewrites.
4The report is printed in the conversation window and automatically saved via save_prompt_evaluation_report.

Error Reference

aiflare.yml not found

aiflare.yml is missing from the project root or api_key is empty.

No session ID

Running outside Claude Code without specifying sessionId, and no .claude-prompts-* file was found. Run inside Claude Code or provide the session ID explicitly.

Session not found

No data exists for the specified session ID. Verify that at least one git commit was made during that session.

Share & Export

Issue a share token or export a file from the report detail page.

Share Link

Share the report via /share/[token] — no authentication required.

Export

Download as Markdown or PDF.

On this page

AIFlare
Made with ♥ in Seoul
© 2026 AIFlare. All rights reserved.
Product
SkillDocs
Company
Vision
Resources
PrivacyTerms
Connect
GitHub