Curated Open-Source AI Repositories

“My RAG ingest front-door — Firecrawl extracts clean markdown where LlamaParse misses tables. Pairs with LlamaIndex Workflows for event-driven retrieval.”

“The Shell skill layer I vendor into Hermes and OpenClaw — 272K stars because it makes Claude Code harness model-visible means logged without rework.”

“My onboarding curriculum for Surat textile founders — 12 Jupyter notebooks that map to our MCP + Pydantic pipeline. When NOT: for prod harness use MAF instead.”

“Runs reachability on our 5-level Claude SDK trees before prod — caught an unbounded loop that would have blown past the 40 brake.”

“Strix typed state rivals LangGraph without the graph verbosity — I prototyped a zero-trust flow in Str ix 40% faster. When NOT: for TS-first still choose Mastra.”

“My terminal baseline for vibe coding 2.0 — I benchmark every harness (98.4% infra paper) against Codex to keep token hunger honest.”

“Graphify holds 0.3% hallucination on our GST cross-check where pgvector alone drifted to 1.2% — the graph is the guardrail.”

“OpenBot governs an agent via AG-UI — I run LangGraph, Mastra or ADK bots as coworkers with separate computers, audit after decision before action.”

“The taste profile skill I plug into Paid Ads Studio — brand consistency without prompt drift across Muse Spark and Grok.”

“I use it to unify security tooling across agents — one skill pack replaces per-agent glue for RE, pentest and CTF; the routing + memory design is why it topped GitHub Trending Aug 1 over openwork and chatwoot. Rating: 4.6/5. When NOT to use: Avoid on production customer data — it is built for authorized lab/CTF scopes, not for compliance-bound pen-tests without isolated VPC.”

“My teams waste hours re-wiring MCP servers per laptop. Openwork solves that — sync workflows/skills in one desktop cowork, which is why it jumped #4→#2 on GitHub Trending Aug 1 alongside reverse-skill. Rating: 4.4/5. When NOT to use: Skip if you need browser-only collaboration — Openwork is desktop-first and shines for local skill/MCP sharing, not pure cloud docs.”

“I curate it over generic agent lists because it gives you a real org chart — frontend, community, reality-checker — each pre-prompted for Claude/Cursor/Codex. The 143k-star flip past semantica on Aug 12 proves the persona model stuck. Rating: 4.8/5. When NOT to use: Not for single-agent loops — overhead is real until you actually run a team-of-agents with shared MCP state.”

“Unlike toy skill demos, Addy’s pack is production-grade — installable via skills CLI into Claude/Cursor with tests. That is why it held steady #4 on Aug 12 while paperclip and orca churned. Rating: 4.7/5. When NOT to use: Not if you want pure prompt tricks — these are opinionated engineering workflows with guardrails, not one-liner hacks.”

“Instead of copying agent prompts, you embed Copilot’s own agentic workflow SDK — April public preview kept it trending Aug 1 at #6 because it turns any app into a Copilot-native agent host. Rating: 4.6/5. When NOT to use: Avoid if you require 100% open-weight inference — the SDK is Copilot-coupled and assumes GitHub-hosted model routing.”

“Video is moving from single-model prompts to 12-pipeline agent teams. OpenMontage packages 700+ skills so a “trailer” spins planner, director and renderer agents — the Aug 12 debut alongside orca makes it the team to watch. Rating: 4.5/5. When NOT to use: Not for quick single-clip Veo calls — OpenMontage pays off for multi-scene, multi-role productions where single prompts fail.”

“I pair Chatwoot with n8n + WhatsApp WABA for SME support automation at ₹27K/mo — full self-hosted control vs Intercom seat tax, which is why it re-trended #7 on Aug 1 amid agent-skill hype. Rating: 4.7/5. When NOT to use: Skip if you need AI-agent-native inbox first — Chatwoot shines for human-led support with automation hooks, not fully autonomous swarms.”