mirzap

9 ms·

151.
by mirzap 1y ago
"with the agent's credentials." - so you are surprised that agent can respond with private repository details when it has access to it? WoW! anyone and anything with credentials can access it. Github action, Jenkins, me. &quo
152.
by mirzap 1y ago
Tomato-Tomato. It's not even an exploit. I will give you my token with access only to public repos. Try and access my private repos with Github MCP. Guess what - you can't - so it is not Github MCP exploit.
153.
by mirzap 1y ago
Of course it will apply. Entire purpose of the agent is to give a response to a prompt. But to sound more dangareous let's call it "injecting". It's a prompt. You are not "injecting" anything. Agent pickups the
154.
by mirzap 1y ago
The role is very simple. It provides an interface for the AI to access the data. Whatever it has access to (via MCP), it will access. Simple as that.
155.
by mirzap 1y ago
It's not even an exploit. MCP is doing what it is MADE TO DO. It's made for interacting with the GitHub API. Whatever it has access to, it will access. If it has access to delete the repo, it will delete the repo. If it has access
156.
by mirzap 1y ago
In principle, I agree with you. The problem I have with articles like this and people commenting is that it's framed as if MCP's vulnerability is discovered, that MCP "needs fixing." When it's not. It's not the
157.
by mirzap 1y ago
@motorest read again what I wrote: "That's the thing. LLM or MCP is not a database. You can’t compare it. You simply can't set the permissions or guardrails within LLMs or MCPs. You always do it layer above (scoping to what L
158.
by mirzap 1y ago
The point is this is NOT a GitHub MCP vulnerability, but how you use it. There is nothing to be fixed in MCP itself; rather how you use it.
159.
by mirzap 1y ago
No, but it's not the database's fault if you don't hash your password. Same here, it's human error, not "MCP vulnerability". It's not that GitHub MCP needs fixing, but rather how you use it. That's th
160.
by mirzap 1y ago
> An attack doesn’t have to be surprising to be an attack. If you open an API to everyone, or put a password as plain text and index it, it's no surprise that someone accesses the "sensitive" data. Nor do I consider that a
161.
by mirzap 1y ago
That's the thing. LLM or MCP is not a database. You can’t compare it. You simply can't set the permissions or guardrails within LLMs or MCPs. You always do it layer above (scoping to what LLM has access to).
162.
by mirzap 1y ago
Would you ever put a plain password text in a search engine and then complain if someone "extracted" that info with a keyword payload?
163.
by mirzap 1y ago
Bad analogy. It's more like indexing a password field in plain text, then opening an API to everyone and setting "guardrails" and permissions on the "password" field. Eventually, someone will extract the data that w
164.
by mirzap 1y ago
I read it, and "attack" does not make sense. If you grant access to MCP to access some data (data you want anybody has access to like public repos, and data that only you want to access to like private repos), you will always be a
165.
by mirzap 1y ago
I simply don't see how you could enforce a classic permission system on an MCP server. MCPs are API servers that allow LLMs access to context within the boundaries you set. You can set permissions for what an LLM has access to and defi
166.
by mirzap 1y ago
I bet you used AI to write something that sounds smart, but trust me kid, it doesn't sound that way at all.
167.
by mirzap 1y ago
How is this considered an "exploit"? You give the agent a token that allows it to access a private repository. MCPs are just API servers. If you don't want something exposed in that API, don't grant them permissions to d
168.
by mirzap 1y ago
Lol. Can't believe I've read this. It's like saying "REST: may cause pwnage". Everything the author found troubling can be said for any API server under certain implementation (or even some frameworks and their defa
169.
by mirzap 2y ago
But that's the point he tries to make. When you "teach" LLM with some knowledge, you teach it a set of patterns. It won't necessarily drop the code that infringes copyright. Say you load Gumroad code into Gemini Pro cont
170.
by mirzap 2y ago
And how exactly are you going to prove derivative work?
171.
by mirzap 2y ago
Let's imagine they do come back. Let's say you have a toaster factory in Vietnam, which allows the US company to produce low-cost toasters for about $100. The factory workers earn $1 per hour. Now, due to the tariffs, a toaster wi
172.
by mirzap 2y ago
Honestly, after playing around with your example and tweaking a JavaScript example, I feel that my plain JS example is faster than the one with WASM/Rust on Nue (possibly due to the animations). Please disable animations; I think they
173.
by mirzap 2y ago
It has nothing to do with architecture, but rather understanding how the DOM works. The DOM is notoriously slow, so you should never render a huge number of rows at once. You can render millions of rows in plain JavaScript without impacting
174.
by mirzap 2y ago
Because he constructs a giant JSON by joining individual entries. Rendering that directly on the DOM will always cause the performance issues (even at the 10k entries). That's why you need to use virtualized list, it can be done in pla
175.
by mirzap 2y ago
Make a demo with react-virtualized[0] and see if it crashes. Hint: It will not[1]. React can easily render 1 million rows with high performance without relying on WASM [2] Here is the demo of react-virtualized[3], in which I entered 10m as
176.
by mirzap 2y ago
Probably. They deny the fact that international commerce is mutually dependent. They think the world strictly depends on them and that companies would not be profitable if they sold products elsewhere. Sure, they will earn less money, but t
177.
by mirzap 2y ago
It's American exceptionalism at its best. They believe America is subsidizing the world when, in fact, it is the other way around. The world is subsidizing the American lifestyle. That's why you see China jumping onto the global s
178.
by mirzap 2y ago
> 150,000 records — far past where JavaScript (and React) would crash with a stack overflow error I think react-virtualized and stack tables can easily handle 1 million rows client-side without a problem (I saw the demo somewhere). Web d
179.
by mirzap 2y ago
Every empire has fallen the same way - spending money on wars and games.
180.
by mirzap 2y ago
I don't think anyone is investing that kind of money without doing basic DD.

More ›