skill-idea-miner

Mine Claude Code session logs into prioritized skill idea candidates.

5|4|Updated Nov 9, 2025
One-click install
npx skills add https://github.com/takusaotome/claude-skills-library --skill skill-idea-miner-takusaotome
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-idea-miner
Source: https://github.com/takusaotome/claude-skills-library/tree/main/skills/skill-idea-miner
Command: npx skills add https://github.com/takusaotome/claude-skills-library --skill skill-idea-miner-takusaotome

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires yaml, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Skill developers waste time manually reviewing Claude Code activity to decide which new skills to build and how to prioritize them.

Core Features & Use Cases

  • Session Log Mining: Scans recent ~/.claude/projects/ JSONL logs, extracts external user requests, detects tool-usage patterns, flags error signals, and finds repetitive manual workflows.
  • Idea Abstraction with LLM Scoring: Uses the Claude CLI to abstract mined signals into candidate skill ideas, then scores novelty, feasibility, and work utility to produce a ranked backlog.
  • Deduplication Against Existing Skills and Backlog: Prevents duplicates by comparing candidates with existing skills/*/SKILL.md frontmatter and the current backlog using Jaccard similarity thresholds.

Quick Start

Run the skill idea generation pipeline with python3 scripts/mine_session_logs.py --dry-run to preview mined candidates without running LLM scoring.

Frequently Asked Questions about skill-idea-miner

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

FAQPage Schema
How do I generate skill ideas from Claude Code session logs?

Skill ideas are generated from Claude Code session logs by extracting external user requests, tool usage signals, and error patterns over a recent lookback window. The pipeline parses JSONL files from `~/.claude/projects/` to identify repetitive workflows and abstracts them into ranked candidates.

How does skill backlog deduplication against existing skills work?

Skill backlog deduplication works by comparing extracted candidate ideas against existing `skills/*/SKILL.md` frontmatter and the current backlog. It uses Jaccard similarity thresholds to prevent overlapping concepts before LLM-based scoring and prioritization.

Can I preview mined skill candidates without running LLM scoring?

Yes, you can preview mined skill candidates without running LLM scoring by executing `python3 scripts/mine_session_logs.py --dry-run`. This outputs extracted workflow patterns and raw signals directly from the session logs for quick review.

What is the best way to prioritize repetitive workflows into new skills?

Prioritizing repetitive workflows into new skills is best handled by LLM-based abstraction that scores novelty, feasibility, and work utility. The pipeline ranks these candidates into a backlog for weekly or on-demand skill topic selection across domains.

Does the session log mining pipeline require YAML dependencies?

Yes, the session log mining pipeline requires the `yaml` dependency to output candidate artifacts for downstream skill generation. These YAML artifacts store the ranked backlog data for further processing and tracking.