
Select between optimjack and 45 other attack scenarios when doing red-teaming. superagent.sh

Making software security autonomous. Free for open source. (YC W24)

Select between optimjack and 45 other attack scenarios when doing red-teaming. superagent.sh

Just released a custom attack scenario: OPTIMJACK Check your front-end for malware injected by thrid-party vendors.

Most red-teaming tools focus on broad vulnerability research. This is very useful but also limited. Today we're releasing ATTACK SCENARIOS so you can hone in on specific threat models that are unique/important for your applications.
Today we're shipping attack scenarios in @superagent_ai . You can now focus repository, application, and agent red teaming on one specific objective. 45 built-in scenarios: 15 for each target type. Things like accessing another tenant's data, escalating privileges, retaining access after revocation, or escaping an execution sandbox.


a security startup raising $25M pre-seed to sell snake oil worries me more than DPRK hackers. false confidence is a hell of an attack surface.

I'm testing a new @superagent_ai feature... We’re tracking threat actor activity: - Who they’re naming - When they’re posting - Which countries and sectors keep appearing I want to follow what each actor is doing and how their activity changes over time. N0n is one example below. Probably DPRK given the geography and types of attacks. Tech behind this is really fun to work with. More soon.


We have the lowest false positive rate in the industry. False positives waste engineering time and erode trust in security tools. Superagent Red Team has ran on over 4500 targets over 500 000 times! - Swarm of hacker agents - Each one with a computer - Each one with an email - Runs exactly like an attacker would Point to your application, infra, agent or repo!

I suffer from Dependabot Fatigue: today I sovled it. I dunno about you but every security PR sends me down the same rabbit hole. Figuring out whether we use the affected functionality. Checking what changed in the update and what might break. @superagent_ai handles it for me!

We're launching Dependency Validation in Superagent. It will save developers countless of hours every day. Today every Dependabot vulnerability advisory becomes a research project. Figuring out whether we use the affected functionality. Checking what changed in the update and what might break. So we built Dependency Validation in Superagent. It investigates Dependabot and Renovate security PRs to check: - whether the vulnerability affects your application - whether the proposed update fixes it - whether the relevant functionality still works Free to use for open-source maintainers and live on the platform: superagent.sh

Forgot to mention this because I was too busy clipping...
Github put Paperclip in the top-10 most-engaged open-source repos for the 2026 alongside Hermes, OpenCode, Pi, and VSCode Pretty good company

I expect the AI labs to absorb a lot of today's security products. I also expect demand for independent security to grow. Everyone is talking about Anthropic's S-1. As a founder in security, I expect the labs to absorb a lot of today's standalone security products. 👇 Three reasons I think independent security has a substantial market ahead:

New milestone reached ⚡️
We just crossed 500 000 PR vulnerability scans across 4125 repositories 👏 @superagent_ai has the lowest FALSE POSITIVE rate compared to OpenAI, Anthropic and CodeRabbit. The reason is our Security Graph. It learns your exact security posture over time across your entire stack.

If you are using Bird here's a tutorial on how to get started: bird.com/ai/superagent

We are teaming up with @messagebird to make messaging workflows safer for AI agents 🤝 Screen inbound email, SMS and WhatsApp messages for prompt injection, phishing and malicious links before your agent acts. Made possible by our security-one our decision model trained on these attacks!
