skill-studio
Design automations from interviews or mine them from your session history.
Interview-driven automation design tool. Runs a coverage-driven JTBD interview (text or voice), then exports a one-page markdown spec plus an SVG design map. Can also analyze Claude Code sessions to extract workflows, subagent patterns, and skill usage, then propose new skills.
What it does
Conducts a structured JTBD interview to help you design a new skill, agent, or automation — then exports a one-page spec and an SVG design map. Can also analyze Claude Code sessions (current or by ID) to extract subagent orchestration, tool sequences, and workflow patterns, then propose new skills based on what it finds.
Key features
- Session analysis — extracts Agent calls, Skill invocations, tool sequences, and repeated workflow patterns from any Claude Code session JSONL
- Current session support — analyze the running session via
--currentwithout needing to look up the ID - Workflow pattern detection — finds repeated multi-tool sequences (e.g. Read → Edit → Bash) that suggest automatable workflows
- Skill proposals — the LLM proposes potential new skills from observed patterns, with rationale citing which signals justify each proposal
- Coverage-driven interview — adapts questions based on which design fields still need data, with sprint/standard/deep depth levels
- Multiple modes — text interview, voice interview, or session-seeded design
When to use
When you want to design a new automation from scratch (interview mode), or when you want to mine an existing session for skill ideas (session analysis mode). Also useful after a productive session to ask “what could be automated from what I just did?”