Pre-built slash-command skills your team can invoke. Each entry covers what it does, when to trigger it, and real enterprise use cases.
Complete agentic software development methodology built on composable skills. Ships 20+ battle-tested skills: TDD, debugging, collaboration, planning. Slash commands include /brainstorm, /write-plan, /execute-plan.
npx skills add obra/superpowers
Install once per team. Establishes a shared methodology so every dev's Claude sessions follow the same structured workflow.
↗ github.com/obra/superpowersCuts ~65% of output tokens by switching Claude to terse caveman mode — drop filler, keep substance, use fragments. Supports lite / full / ultra intensity levels. Works as Claude Code plugin, Codex plugin, Cursor rule, and 40+ other agents.
/caveman · /caveman lite|full|ultranpx skills add JuliusBrussee/caveman
Activate on long sessions or when context window is filling fast. 65% fewer output tokens = meaningfully lower cost and faster responses.
↗ github.com/JuliusBrussee/cavemanPre-indexed code knowledge graph that auto-syncs on file changes. Exposes symbol relationships, call graphs, and code structure as an MCP server — agents query the graph instead of scanning files. 100% local, zero file contents leave the machine.
Auto-configured MCP server · no slash command needednpx codegraph install
Large codebases where Claude repeatedly scans the same files. Wires into Claude Code, Cursor, Codex, Gemini CLI, and Kiro automatically on first run.
↗ github.com/colbymchenry/codegraphTurns any folder of code, SQL schemas, docs, papers, or images into a queryable knowledge graph. Supports 20+ languages via tree-sitter AST (Python, TS, Go, Rust, Java, C, C++, Ruby, C# and more). All processing is local — no file contents leave your machine.
/graphifynpm install -g @sentropic/graphify && graphify install
Map an unfamiliar codebase in minutes. Also useful for combining app code + database schema + infrastructure into one queryable graph at project start.
↗ github.com/safishamsi/graphifySingle Rust binary that sits between Claude Code and your shell. Compresses git diff, cargo test, grep, find, and 30+ other commands before output hits the context window — 60–90% token savings with zero config changes. Wires in via a PreToolUse hook so every Bash call is transparently rewritten.
rtk init -g # after installing binary from rtk-ai.app
Activate once per machine. Shaves 70–90% off long sessions where Claude repeatedly reads test output, diffs, or directory listings — meaningfully lower cost and faster context turnaround.
↗ github.com/rtk-ai/rtkForces the laziest solution that actually works. YAGNI-first escalation ladder: does this need to exist → stdlib → native platform feature → already-installed dep → one line → only then the minimum viable implementation. Marks every shortcut with a ponytail: comment naming its upgrade path, so deferred work becomes a ledger, not amnesia.
npx skills add dietrichgebert/ponytail
Eliminate scope creep and over-engineering during feature work. Pairs well with superpowers for teams that want both a structured workflow and a "ship the simplest thing" discipline.
↗ github.com/dietrichgebert/ponytail