CrabTrap: An LLM-as-a-judge HTTP proxy to secure agents in production
106 points
14 hours ago
| 11 comments
| brex.com
| HN
https://www.brex.com/journal/building-crabtrap-open-source
simonw
3 hours ago
[-]
Comments like this don't fill me with confidence: https://github.com/brexhq/CrabTrap/blob/4fbbda9ca00055c1554a...

  // The policy is embedded as a JSON-escaped value inside a structured JSON object.
  // This prevents prompt injection via policy content — any special characters,
  // delimiters, or instruction-like text in the policy are safely escaped by
  // json.Marshal rather than concatenated as raw text.
reply
yakkomajuri
6 hours ago
[-]
Really cool! I'm also building something in this space but taking a slightly different approach. I'm glad to see more focus on security for production agentic workflows though, as I think we don't talk about it enough when it comes to claws and other autonomous agents.

I think you're spot on with the fact that it's so far it's been either all or nothing. You either give an agent a lot of access and it's really powerful but proportionally dangerous or you lock it down so much that it's no longer useful.

I like a lot of the ideas you show here, but I also worry that LLM-as-a-judge is fundamentally a probabilistic guardrail that is inherently limited. How do you see this? It feels dangerous to rely on a security system that's not based on hard limitations but rather probabilities?

reply
manapause
36 minutes ago
[-]
Correct me if I’m wrong, but from my experience in this space in order for a model to exercise judgment it must force itself to operate in a strict chain of thought mode. Since all LLMs are predictive creatures, I started to care a lot more about my judgment settings, the transparency of them, and the presence of a judgment loop in either the development or functionality of an application built these days.

Not exactly sure where I’m going with this, but my work with creating penetesting tools for LLMs, the way that I use judgment is critical to the core functionality of the application. I agree with your concern and I will just say that the more time I spent concerned with chain of though where now I will make multiple versions of the same app using a different judge set a different “temperaments” and I found it to be incredibly enlightening as to the diversity of applications and approaches that it creates.

  Even using BMAD or superpowers, I can make five versions of an app without judges involved and I feel like I’m just making the same app five times because the API begins to coalesce around the business problem you want to solve. The vicissitudes of prediction tools always want to take the safest bet for the greater good, but with the judge involved we can make the agent force itself to actually be hostile about what exactly we’re trying to do, which has produced interesting and fun results.
reply
roywiggins
5 hours ago
[-]
It's all fine until OpenClaw decides to start prompt injecting the judge
reply
bambax
4 hours ago
[-]
Exactly; would probably be safer with a purely algorithmic decision making system.
reply
fc417fc802
4 hours ago
[-]
Calling it now. Show HN: Pincer - A small highly optimized local model to detect prompt injection attempts against other models.
reply
reassess_blind
3 hours ago
[-]
Sounds like a good idea. Please send me the Github link once done and I'll have my OpenClaw take a look and form my opinion of it.
reply
NamlchakKhandro
50 minutes ago
[-]
Sounds like a good idea. Please send me you GitHub now and I'll have my big claw crush your open claw
reply
foreman_
47 minutes ago
[-]
The thread has converged on “LLM-as-judge is the wrong security primitive,” which is right as far as it goes. The prompt-injection chain ends at the outbound POST. By the time the judge sees the request, the credential has already been read.

The question edf13 pointed at but didn’t develop; where does a transport-layer judge earn its place at all? Not as the enforcement layer but as the audit layer on top of one. Kernel-level controls tell you what the agent did. A proxy tells you what the agent tried to exfiltrate and where to.

Structured-JSON escaping and header caps are good tools for the detection job. They’re the wrong tools for the prevention job. Different layers, different questions.

reply
edf13
1 hour ago
[-]
Good to see more work in this space with different ideas. The policy-builder-from-traffic idea is genuinely novel.

We looked at LLM-as-judge early on and ended up discounting it on security grounds: the judge itself sits in the prompt-injection blast radius, and a probabilistic gate protecting a probabilistic agent felt like the wrong shape for a security primitive. Their structured-JSON escaping and header/body caps are thoughtful mitigations, but they reduce the surface rather than eliminate it.

Picking the transport layer makes sense for production API-calling agents where egress is where irreversible damage lands. The architectural tradeoff is what the proxy can't see: file reads, shell spawns, process execs. The canonical prompt-injection chain (malicious README -> read ~/.ssh/id_rsa -> POST to attacker.com) is three steps, and CrabTrap only sees step three. The credential has already left the filesystem and entered agent process memory by the time the judge evaluates the outbound request.

HTTP_PROXY/HTTPS_PROXY also depends on cooperative libraries. The iptables note handles this well in a containerised production deploy. For local-laptop coding agents, which is where most prompt-injection attack surface lives today, there's no equivalent kernel-level backstop.

For that threat model we've been building grith.ai at the syscall layer (ptrace/seccomp-BPF on Linux, Endpoint Security on macOS, Minifilter + ETW on Windows) rather than transport. The two compose cleanly; serious production deploys probably want both.

reply
lmeyerov
1 hour ago
[-]
At RSAC, there were a ton of agentic security startups converging on ebpf monitors for this reason. Eg, sondera gave a fun talk at graph the planet where they did that + exposed with a policy layer over agent traces via Cedar (used in AWS IAM etc). ABAC and identity were also appearing near here.

One thing I didn't see: are there any OSS solutions appearing here?

reply
edf13
1 hour ago
[-]
We are Open Source… code will be published soon (before launch)
reply
lanyard-textile
41 minutes ago
[-]
Then you will be open source ;) Not yet open source.
reply
babas03
2 hours ago
[-]
The LLM-as-judge approach keeps coming up (some agent platforms use a dual-LLM validator; there's active research around it) and I'm curious how CrabTrap handles the latency-vs-safety tradeoff. Does the judge run on every call, or only on calls that trip a deterministic policy first? In the payments/ads domain specifically, the blast radius of a mis-approved call is high enough that "another LLM says OK" can feel like trading one black box for two.

Also interesting that you went HTTP. Most agent tooling I've been running is stdio-based (MCP-style). What did the HTTP framing buy you architecturally?

Why it lands: specific technical question, credits their work, ends with something that invites response. If Brex engineers are in the thread, one of them will likely reply.

reply
fareesh
3 hours ago
[-]
Needs to be deterministic. ACLs
reply
erdaniels
3 hours ago
[-]
Yes, full stop. They say they cap the body to 16k and give the LLM a warning, lol. And this is coming from a credit card company.
reply
ArielTM
2 hours ago
[-]
The debate here is missing a practical question: is the judge from the same model family as the agent it's judging?

If both are Claude, you have shared-vulnerability risk. Prompt-injection patterns that work against one often work against the other. Basic defense in depth says they should at least be different providers, ideally different architectures.

Secondary issue: the judge only sees what's in the HTTP body. Someone who can shape the request (via agent input) can shape the judge's context window too. That's a different failure mode than "judge gets tricked by clever prompting." It's "judge is starved of the signals it would need to spot the trick."

reply
IntrepidPig
1 hour ago
[-]
Blatant “astroturfing” in these comments
reply
Seventeen18
4 hours ago
[-]
So cool ! I'm building something very close to that but from another perspective, making this open source is giving me many idea !
reply
DANmode
6 hours ago
[-]
We’re supposed to be fixing LLM security by adding a non-LLM layer to it,

not adding LLM layers to stuff to make them inherently less secure.

This will be a neat concept for the types of tools that come after the present iteration of LLMs.

Unless I’m sorely mistaken.

reply
reassess_blind
6 hours ago
[-]
It looks as if this tool has traditional static rules to allow/deny requests, as well as a secondary LLM-as-a-judge layer for, I imagine, the kinds of rules that would be messy or too convoluted to implement using standard rules.
reply
stingraycharles
3 hours ago
[-]
I think the parent’s point is that this should be implemented using e.g. Bayesian statistics rather than an LLM, as the judge LLM is vulnerable to the exact same types of attacks that it’s trying to protect against.

Most proper LLM guardrails products use both.

reply
snug
6 hours ago
[-]
I think this can be great as additional layer of security. Where you can have a non llm layer do some analysis with some static rules and then if something might seem phishy run it through the llm judge so that you don’t have to run every request through it, which would be very expensive.

Edit: actually looks like it has two policy engines embedded

reply
windexh8er
6 hours ago
[-]
And we don't think the judge can/will be gamed? Also... It's an LLM, it's going to add delay and additional token burn. One subjective black box protecting another subjective black box. I mean, what couldn't go wrong?
reply
ImPostingOnHN
6 hours ago
[-]
What happens when a prompt injection attack exploits the judge LLM and results in a higher level of attacker control than if it never existed?
reply
vova_hn2
5 hours ago
[-]
How can it result in a higher level of control? I don't see why the "judge" should have access to anything except one tool that allows it to send an "accept" or "deny" command.
reply
nl
5 hours ago
[-]
> We’re supposed to be fixing LLM security by adding a non-LLM layer to it,

If people said "we build a ML-based classifier into our proxy to block dangerous requests" would it be better? Why does the fact the classifier is a LLM make it somehow worse?

reply
Retr0id
5 hours ago
[-]
The fact that LLMs are "smarter" is also their weakness. An oldschool classifier is far from foolproof, but you won't get past it by telling it about your grandma's bedtime story routine.
reply
reassess_blind
3 hours ago
[-]
Fairly hard to bypass the latest LLMs with grandma's bedtime story these days, to be fair.
reply
Retr0id
3 hours ago
[-]
That specific trick yes, but the general concept still applies.
reply
reassess_blind
3 hours ago
[-]
It does, but it's certainly not trivial. In fact there's an unclaimed $1000 bounty on prompt injecting OpenClaw: https://hackmyclaw.com/
reply
DANmode
1 hour ago
[-]
Is that enough?
reply
waterTanuki
5 hours ago
[-]
If you're working in a mission-critical field like healthcare, defense, etc. you need a way to make static and verifiable guarantees that you can't leak patient data, fighter jet details etc. through your software. This is either mandated by law or in your contract details.

The entire purpose of LLMs is to be non-static: they have no deterministic output and can't be validated the same way a non-LLM function can be. Adding another LLM layer is just adding another layer of swiss cheese and praying the holes don't line up. You have no way of predicting ahead of time whether or not they will.

You might say this hasn't prevented leaks/CVEs in exisiting mission-critical software and this would be correct. However, the people writing the checks do not care. You get paid as long as you follow the spec provided. How then, in a world which demands rigorous proof do you fit in an LLM judge?

reply
nl
2 hours ago
[-]
> The entire purpose of LLMs is to be non-static: they have no deterministic output and can't be validated the same way a non-LLM function can be. Adding another LLM layer is just adding another layer of swiss cheese and praying the holes don't line up. You have no way of predicting ahead of time whether or not they will.

This is exactly the point though. A LLM is great at finding work-around for static defenses. We need something that understands the intent and responds to that.

Static rules are insufficient

reply
SkyPuncher
6 hours ago
[-]
Defense in depth. Layers don't inherently make something less secure. Often, they make it more secure.
reply
yakkomajuri
6 hours ago
[-]
I do think this is likely to make things more secure but it's also dangerous by potentially giving users a false sense of complete security when the security layer is probabilistic rather than deterministic.

EDIT: it does seem to have a deterministic layer too and I think that's great

reply