Skip to content

Skills Overview

Distillery provides 10 Claude Code slash commands organized into three categories: knowledge capture, knowledge retrieval, and ambient intelligence.

Knowledge Capture

Skill Purpose
/distill Capture decisions, insights, and action items from working sessions
/bookmark Save URLs with auto-generated summaries
/minutes Create and update structured meeting notes

Knowledge Retrieval

Skill Purpose
/recall Semantic search over the knowledge base
/pour Multi-entry synthesis with citations and gap analysis
/classify Classify entries by type and triage the review queue

Ambient Intelligence

Skill Purpose
/watch Manage monitored feed sources (GitHub, RSS)
/radar Generate a digest of recent feed activity
/tune Adjust feed relevance thresholds

Onboarding

Skill Purpose
/setup First-time configuration wizard

How Skills Work

Skills are defined as SKILL.md files in .claude-plugin/skills/. When you invoke a slash command, Claude Code loads the markdown and follows the instructions — no Python code is executed by the skill itself. All data operations go through the Distillery MCP server's tools.

Common Patterns

All skills share these conventions (defined in CONVENTIONS.md):

  • MCP health check — skills verify the MCP server is reachable before proceeding
  • Author resolution — determined from git config user.name, then DISTILLERY_AUTHOR env var, then asks
  • Project resolution — determined from --project flag, then git rev-parse --show-toplevel, then asks
  • Tag extraction — 2-5 keywords, lowercase, hyphen-separated, hierarchical (e.g., project/distillery/decisions)
  • Error handling — display the error and stop; no retry loops
  • Preview before store — all write operations show a preview and ask for confirmation