DEV Community

Tang Weigang profile picture

Tang Weigang

Independent AI capability pack builder. Doramagic resource packs are not official upstream projects.

Joined Joined on  Personal website https://doramagic.ai
An MCP Server Is Not Safe Just Because It Started

An MCP Server Is Not Safe Just Because It Started

Comments
1 min read
Your AI Coding Policy Needs a Receipt, Not Just an AGENTS.md

Your AI Coding Policy Needs a Receipt, Not Just an AGENTS.md

Comments
2 min read
agent-workspace-linux Is Not Computer Use: Verify the Workspace Boundary First

agent-workspace-linux Is Not Computer Use: Verify the Workspace Boundary First

Comments
1 min read
Sverklo Is Not Just Grep with Embeddings: Verify the Code-Memory Evidence Chain

Sverklo Is Not Just Grep with Embeddings: Verify the Code-Memory Evidence Chain

Comments 1
3 min read
OpenViking Is Not Just a Memory Folder: Verify Hierarchical Context Delivery First

OpenViking Is Not Just a Memory Folder: Verify Hierarchical Context Delivery First

Comments
2 min read
Notion MCP: Verify Page Access Before You Trust the Tool List

Notion MCP: Verify Page Access Before You Trust the Tool List

Comments
2 min read
Graphiti Is Not Just Another RAG Layer: Keep Time and Provenance

Graphiti Is Not Just Another RAG Layer: Keep Time and Provenance

Comments
2 min read
FastMCP: tool discovery is not authorization

FastMCP: tool discovery is not authorization

Comments
1 min read
Before an AI agent edits a repo, give it evidence it can audit

Before an AI agent edits a repo, give it evidence it can audit

Comments
1 min read
Before an AI agent touches a repo, give it a verifiable context pack

Before an AI agent touches a repo, give it a verifiable context pack

Comments
2 min read
Before Adopting OpenAI Agents SDK, Write the Runtime Contract First

Before Adopting OpenAI Agents SDK, Write the Runtime Contract First

Comments
5 min read
Before Adopting mem0, Decide What an Agent Is Allowed to Remember

Before Adopting mem0, Decide What an Agent Is Allowed to Remember

Comments
5 min read
Before adopting DSPy, prove the LM program has a contract

Before adopting DSPy, prove the LM program has a contract

Comments
3 min read
LlamaIndex Is Not a Five-Line RAG Demo. First Prove the Context Contract.

LlamaIndex Is Not a Five-Line RAG Demo. First Prove the Context Contract.

Comments
6 min read
GPT Researcher Is a Source Pipeline. First Prove the Citation Contract.

GPT Researcher Is a Source Pipeline. First Prove the Citation Contract.

Comments
5 min read
Gemini CLI Brings Gemini Into The Terminal. First Prove The Tool Boundary.

Gemini CLI Brings Gemini Into The Terminal. First Prove The Tool Boundary.

Comments
4 min read
OpenHands Can Run Coding Agents. First Decide What They Are Allowed to Touch.

OpenHands Can Run Coding Agents. First Decide What They Are Allowed to Touch.

Comments
4 min read
Serena Gives Agents Semantic Code Tools. First, Prove the Project Boundary.

Serena Gives Agents Semantic Code Tools. First, Prove the Project Boundary.

Comments
4 min read
FastMCP Makes MCP Easy. The First Risk Is Exposing Too Much.

FastMCP Makes MCP Easy. The First Risk Is Exposing Too Much.

Comments
3 min read
smolagents Is Powerful Because It Runs Code. That Is Also the Boundary.

smolagents Is Powerful Because It Runs Code. That Is Also the Boundary.

Comments
4 min read
Browser Agents Need Receipts, Not Just Clicks: A Practical Read of browser-use

Browser Agents Need Receipts, Not Just Clicks: A Practical Read of browser-use

Comments
4 min read
Do Not Treat Pydantic AI as an Agent Magic Layer

Do Not Treat Pydantic AI as an Agent Magic Layer

Comments
5 min read
Before You Ship an Agent, Make DeepEval Test the Failure Path

Before You Ship an Agent, Make DeepEval Test the Failure Path

Comments
5 min read
Before an Agent Uses Qdrant, Write the Retrieval Contract

Before an Agent Uses Qdrant, Write the Retrieval Contract

Comments
4 min read
Before an Agent Runs Code in E2B, Define the Sandbox Contract First

Before an Agent Runs Code in E2B, Define the Sandbox Contract First

Comments
4 min read
Do not treat LangGraph as a longer chain: define state, interrupts, and recovery first

Do not treat LangGraph as a longer chain: define state, interrupts, and recovery first

Comments
4 min read
Do not add LLM evals after launch: use promptfoo to define the failure boundary first

Do not add LLM evals after launch: use promptfoo to define the failure boundary first

Comments
4 min read
Before You Let an Agent Convert Documents, Narrow the MarkItDown Boundary

Before You Let an Agent Convert Documents, Narrow the MarkItDown Boundary

Comments
4 min read
Before You Add Memory to an AI Agent, Decide What the Agent Is Allowed to Remember

Before You Add Memory to an AI Agent, Decide What the Agent Is Allowed to Remember

Comments
5 min read
Before You Give an AI Agent a Browser, Define the Puppeteer Boundary

Before You Give an AI Agent a Browser, Define the Puppeteer Boundary

Comments
4 min read
Before You Trust an LLM App, Review the Trace

Before You Trust an LLM App, Review the Trace

Comments
4 min read
Before You Add More Agents, Design the Control Plane

Before You Add More Agents, Design the Control Plane

Comments
5 min read
Adopt Codex CLI only after you can explain the source, boundary, review, and rollback model

Adopt Codex CLI only after you can explain the source, boundary, review, and rollback model

Comments
5 min read
Codex CLI is useful only when the workflow around it is reviewable and reversible

Codex CLI is useful only when the workflow around it is reviewable and reversible

Comments
4 min read
Codex CLI Is Useful Only When the Workflow Around It Is Reviewable and Reversible

Codex CLI Is Useful Only When the Workflow Around It Is Reviewable and Reversible

Comments
7 min read
Codex CLI needs a rollback-ready workflow, not just a quick install

Codex CLI needs a rollback-ready workflow, not just a quick install

Comments
1 min read
Aider clone signals point to rollback-ready coding contracts, not just better code generation

Aider clone signals point to rollback-ready coding contracts, not just better code generation

Comments
4 min read
Chrome DevTools MCP traffic shows developers are checking operability, not hype

Chrome DevTools MCP traffic shows developers are checking operability, not hype

Comments
3 min read
A clone-heavy MCP pack signal is about inspectability, not marketing

A clone-heavy MCP pack signal is about inspectability, not marketing

Comments
2 min read
Complex AI frameworks need acceptance-ready context packs, not longer prompts

Complex AI frameworks need acceptance-ready context packs, not longer prompts

Comments
3 min read
AI skills should be portable capability assets, not prompt collections

AI skills should be portable capability assets, not prompt collections

Comments
2 min read
AI coding agents need browser evidence, not just code inspection

AI coding agents need browser evidence, not just code inspection

Comments
1 min read
Load OpenSkills context into Codex or Claude Code without guessing source boundaries

Load OpenSkills context into Codex or Claude Code without guessing source boundaries

Comments
1 min read
How to tell whether an AI capability pack can actually help you ship

How to tell whether an AI capability pack can actually help you ship

Comments
2 min read
Publishing AI capability packs is not a content blast

Publishing AI capability packs is not a content blast

Comments
2 min read
An AI agent resource pack is more than a prompt

An AI agent resource pack is more than a prompt

2
Comments
3 min read
loading...