Inner-loop coding helpers: skills to organize uncommitted changes into clean commits (git-commit) and rewrite messy branch history before a PR (rewrite-commits), plus subagents for codebase research, runtime verification, test running, commit organization, and scope guarding.
2 skills · 5 agents
Distill insights from browser bookmarks/history, Pocket/Raindrop exports, or URL lists into an Obsidian vault. Scales to 76K+ URLs via domain triage, curl-based scraping, and model routing (haiku for triage, sonnet for classification).
1 skill · 3 commands
Write lean, effective agent instructions — context files (AGENTS.md, CLAUDE.md) and slash commands. Most auto-generated ones make agents worse (arXiv:2602.11988v1). This skill knows what to keep and what to cut.
1 skill · 2 commands
Multi-agent orchestration: spin up a read-only research team to explore a topic from multiple angles, or an implementation team that follows a phased PRAR workflow.
2 skills
Quality control for written output: flag AI-slop in agent-generated text and suggest rewrites, and audit documents for unverified technical claims, verifying attributions against real sources (SIFT).
1 skill · 1 agent
Standalone utility skills: look at local images with a local vision model (OCR, describe, read diagrams) and prune stale ~/.claude/projects/ folders.
2 skills
Generate a learnxinyminutes.com-style tutorial for anything: a CLI command, language, library, framework, algorithm, concept, or GitHub repo.
1 skill