Local Skill Generation (No Account Required)
Generate reusable Claude Code slash commands from your existing AI coding transcripts — entirely offline. Free forever, no signup, no data leaves your machine.
How it works
- PromptConduit reads transcripts from Claude Code (
~/.claude/projects/*.jsonl), OpenAI Codex CLI, and GitHub Copilot CLI.
- It detects repeated workflows — sequences of prompts and tool calls you've used multiple times.
- It writes each detected pattern as a Claude Code skill file in
~/.claude/skills/.
- The skill is immediately available as a slash command in any new Claude Code session.
Generate skills
promptconduit skills generate --local
Add --dry-run to preview what would be written without touching disk.
Supported transcript sources
- Claude Code:
~/.claude/projects/*/sessions/*.jsonl
- OpenAI Codex CLI:
~/.codex/history/*.json
- GitHub Copilot CLI:
~/.copilot/sessions/*.jsonl
Privacy
Everything runs locally. No transcripts, prompts, or detected patterns are sent to any server when using --local. This is the recommended mode for individual developers and privacy-sensitive teams.
Other Documentation
- Installation — Install the PromptConduit CLI via Homebrew, curl, or from source. Available for macOS, Linux, and Windows.
- Supported Tools — PromptConduit supports Claude Code, Cursor, Gemini CLI, Codex, and Copilot for real-time capture and skill generation.
- Configuration — Configure the PromptConduit CLI: API keys, environments, debug mode, and environment variables.
- CLI Reference — Full CLI reference for PromptConduit. Commands for install, sync, skills generation, config, and more.
- API Reference — PromptConduit REST API reference. Ingest raw events via /v1/events/raw with RFC 7807 error responses.