Repository navigation
Releases: 4pixeltechBR/promptograph
Releases · 4pixeltechBR/promptograph
Release list
📷 Promptograph v0.3.2 - Official PyPI Package & Zero-Install uvx
📷 Promptograph v0.3.2 — Official PyPI Release & Zero-Install uvx Support
Release date: 2026-09-27
PyPI Package: https://pypi.org/project/promptograph/
🚀 Highlights
- 📦 Published on PyPI: Zero-install execution worldwide via
uvx promptograph. - 🔌 FastMCP Server Ready: Automatic stdio MCP server for Claude Desktop, Cursor, Windsurf, and autonomous agents.
- 📚 20,475 Real Prompts: Full corpus embedded and compressed (~660 KB).
- 🧩 226 Curated Agent Skills: Instant access to 37 engineering, multimedia, and automation domains.
⚡ Quick Connect (Claude Desktop / Cursor)
{
"mcpServers": {
"promptograph": {
"command": "uvx",
"args": ["promptograph"]
}
}
}📷 Promptograph v0.3.1 - The Agent Intelligence & Skills Release
📷 Promptograph v0.3.1 — "The Agent Intelligence & Skills Release"
Release date: 2026-09-25
Tagline: Photograph every system prompt that matters & Equip AI Agents with Curated Skills.
🎯 Highlights
- 🔌 Native FastMCP Server: 7 operational tools over standard I/O (stdio), instantly connecting Claude Desktop, Claude Code, Cursor, Windsurf, and Antigravity to your prompt & skills intelligence.
- 🧩 Curated Agent Skills Registry: 226+ operational skills blueprints across 37 engineering, AI, multimedia, and automation domains (
data/skills_manifest.json~380 KB). - 📚 Frontier Corpus Integration: Full parity with the 20,475 real-world system prompts and ~94 million tokens index.
- 🛡️ Strict ToS Compliance & Secret Shielding: 100% audited against GitHub Acceptable Use Policies; zero leaked tokens, zero attack exploits, zero personal machine paths.
- 🚀 Zero Heavy Dependencies: Lightweight Python 3.10+ standard library implementation with FastMCP support.
🆕 What's New in v0.3.1
1. 🔌 FastMCP Server for Autonomous Agents
Agents no longer need to parse raw files. Connect Promptograph directly as an MCP server:
{
"mcpServers": {
"promptograph": {
"command": "python",
"args": ["scripts/promptograph_mcp_server.py"]
}
}
}2. 🛠️ The 7-Tool Agent Suite
| Tool | Domain | Purpose |
|---|---|---|
promptograph_search |
Prompts | Search 20,475 real prompts by company, model, or textual query |
promptograph_validate |
Heuristics | Deterministic 13-rule audit scoring prompts (0-100%, Grade A+ to F) |
promptograph_generate |
Synthesis | Generate battle-tested prompts using presets (Claude, GPT, Cursor, Devin) |
promptograph_stats |
Metrics | Query corpus metrics, token counts, and laboratory breakdowns |
promptograph_skills_search |
Skills | Discover 226+ curated skills across 37 technical domains |
promptograph_skills_get |
Skills | Retrieve executable operational blueprints (SKILL.md) dynamically |
promptograph_skills_categories |
Skills | Get domain breakdown and catalog distributions |
3. 🧩 Curated Skills Domains (226+ Blueprints)
Includes operational recipes for:
- Agents & Multi-Agent Swarms: Kokoro Realtime Voice, Autonomous Coding Agents, OmniParser GUI
- Elite Software Engineering: Clean Code, Architecture Design, Profiling, Refactoring
- Media & 3D Synthesis: ComfyUI Production Nodes, Remotion Video Pipelines, 3DGS
- Quantitative Modeling: DeFi algorithmic models, signal normalizers, backtesting recipes
- Web & Local Delivery: Lumora Studio, Brazilian Local Business LPs, Adaptive Crawlers
📊 Final Stats (v0.3.1)
| Metric | Value |
|---|---|
| Prompts indexed | 20,475 |
| Curated Skills indexed | 226+ |
| Skill Domains | 37 categories |
| FastMCP Tools | 7 active tools |
| Tokens indexed | ~94,000,000 |
| Companies | 40+ (Anthropic, OpenAI, Google, xAI, Meta, Mistral, DeepSeek) |
| Backend footprint | ~50 MB RAM |
🔗 Links
- 🌐 Repository: https://github.com/4pixeltechBR/promptograph
- 📊 Live Site: https://4pixeltechBR.github.io/promptograph/
- 📜 License: MIT © 2026 4Pixel Tech
📷 Promptograph v0.3.0 - The Frontier Update
📷 Promptograph v0.3.0 — "The Frontier Update"
Release date: 2026-09-03
Tagline: Photograph every system prompt that matters.
🎯 Highlights
- 🚀 20,475 prompts indexed (was 10,483 in v0.2.0 — +95%!)
- 📦 1.3GB archive (was 802MB — +62%)
- 📚 55 repos (was 49)
- 🔤 70.5M words indexed
- 🆕 All major official open source projects integrated
🆕 What changed since v0.2.0
🏗️ Official open source integrations (THE BIG ONE)
| Repo | Stars | Why it matters |
|---|---|---|
deepseek-ai/deepseek-harness |
⭐ 210.9k | Full agent framework, MIT, "Everything is a Plugin" |
ultraworkers/claw-code |
⭐ 195.2k | Fastest GitHub repo to 100k stars — Rust rewrite of Claude Code |
anthropics/claude-code |
⭐ 143.9k | Official Claude Code source |
openai/codex |
⭐ 121.2k | Official OpenAI Codex |
alchaincyf/deepseek-orange-book |
⭐ 1.2k | 120-page reverse-engineering book |
bradAGI/awesome-cli-coding-agents |
⭐ 1.1k | Curated list of 30+ CLI agents |
🆕 New system prompts (Aug-Sep 2026)
Anthropic:
- Claude Cowork + dispatch (246KB combined)
- Claude Science (155KB)
- Claude Design (142KB, 53 tools, 22 skills)
- Claude in Chrome (74KB)
- Claude for Word/Excel/PowerPoint (51KB combined)
- Voice Mode, Interviewer, Reminders, Visualize
OpenAI:
- GPT-5.6 Sol, Terra, Luna (separated)
- Codex GPT-5.6 Sol variant
xAI (SpaceXAI):
- Grok 4.5, Grok 4.6, Grok Bot
Google:
- Gemini 3.7 Flash
Meta:
- Muse Code
Moonshot:
- Kimi K2.6 (1T params)
Z.ai (Zhipu):
- ZCode / GLM-5.3 (391KB combined — 3 files)
- Ox Alpha / GLM-5.3-Flash (stealth leak)
Coding agents:
- OpenCode, Pi, CommandCode CLI
- CodeBuddy, Warp.dev, Trae, Kiro, Qoder, Leap.new
🔐 Security research artifacts
- CVE-2026-54316 (Claude Code side-channel)
- CVE-2026-12537 (Gemini CLI, CVSS 10.0)
- CVE-2026-21852 (Claude Code pre-trust leak)
- CVE-2026-24301 (CoSnitch Copilot Personal)
- "Stolen Thoughts" attack (arXiv:2608.09867)
- Autonomous operation prompt reveal
🛠️ Stack (unchanged)
- Python 3.10+ stdlib (zero dependencies)
- HTML/JS vanilla (sem npm, sem build)
- MIT license
🔗 Links
- 🌐 Site: https://4pixeltechBR.github.io/promptograph/
- 💻 Repo: https://github.com/4pixeltechBR/promptograph
- 📚 Archive: 1.3GB, 55 repos, 20,475 prompts
📊 Stats finais
| Metric | Value |
|---|---|
| Prompts | 20,475 |
| Words | 70.5M |
| Lines | 4.7M |
| Repos | 55 (+ 6 HF datasets) |
| Companies | 40+ |
| Archive size | 1.3GB |
| Date range | 2022 — Sep 2026 |
Comparação v0.2 → v0.3:
- Prompts: 10,483 → 20,475 (+95%)
- Repos: 49 → 55 (+12%)
- Words: 13.7M → 70.5M (+415%)
- Archive: 802MB → 1.3GB (+62%)