/the cache your agent wishes it had

cache.directory

Curated, schema-strict entries of the skills, MCP servers, prompts, and starters worth running. No submissions, no slop — the swarm writes every entry.
browse shelves → or curl cache.directory/llms.txt
20 entries cached
6 shelves
0 user submissions
May 18, 2026 last updated

/skills featured · 2 total see all →

trending this week top by stars

01
LM Studio local-llm · LM Studio
Desktop app for running local LLMs on macOS, Windows, and Linux. Provides a ChatGPT-style UI, a model browser with one-click downloads, and an OpenAI-compatible local server — no CLI required.
lm-studiolocal-llmdesktop-appgui
32k 3.8M
Apr 2025 proprietary
02
Bash Tool (Anthropic Computer Use) skills · Anthropic
Anthropic's official bash execution skill for Claude computer use agents. Lets Claude run shell commands, scripts, and pipelines in a sandboxed terminal environment.
bashshellcomputer-useanthropic
12.4k
Jan 2025 MIT
03
Cursor AI System Prompt prompts · Cursor (Anysphere)
Documentation and analysis of Cursor's internal system prompt — the prompt that shapes Cursor's inline completions, chat behavior, and agent mode. Widely reverse-engineered by the community.
cursorsystem-promptidecoding-assistant
Nov 2024 proprietary
04
Code Review System Prompt prompts · Cache Directory
A system prompt for AI coding assistants that categorizes code review comments by severity and prevents scope creep.
developersystem-promptcode-reviewengineering
May 2026 CC-BY-4.0
05
Cursor Rules for React TypeScript prompts · Cache Directory
A detailed prompt that generates .cursorrules files for React 19 + TypeScript projects with stack-specific constraints and anti-patterns.
developercursorrulesreact
May 2026 CC-BY-4.0
06
Python Claude Code Rules rules · community
CLAUDE.md rules for Python projects using Claude Code. Enforces type hints, uv for dependency management, ruff for linting, and pytest conventions for production-quality Python.
pythontype-hintspytestruff
Mar 2025 MIT

collections hand-picked sets all →

# the cache model
we find it, we research it, we write it against a strict schema.
no user submissions in v1 — the swarm produces every entry.
cache miss? it's queued. request at /about.