Adversarial security testing for AI agents — prompt injection, tool misuse, memory poisoning, and MCP scanning, with CI-ready regression testing and reports.
How fast invarislabs/invaris-agentsec is gaining stars, and where it lands on Trendshift and GitHub Trending
Activity and trending
Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending
Oct 29 – Nov 5, 2026
Stars +4in 8 days
012
2
Forks +0in 8 days
01
Merged PRs 2in 8 days
01
1
Trendshift Trendingwhere the repository ranked each day it made the top 25
Not in the Trendshift top 25 in this period
Oct 29Oct 30Oct 31Nov 1Nov 2Nov 3Nov 4Nov 5
The left and right arrow keys step from day to day; Home and End go to the first and the last
Stars, forks, merged pull requests and Trendshift Trending rank, by day
Day
Stars gained
Forks gained
Merged PRs
Thu, Oct 29, 2026
+0
+0
0
Fri, Oct 30, 2026
+1
+0
1
Sat, Oct 31, 2026
+2
+0
1
Sun, Nov 1, 2026
+0
+0
0
Mon, Nov 2, 2026
+1
+0
0
Tue, Nov 3, 2026
+0
+0
0
Wed, Nov 4, 2026
+0
+0
0
Thu, Nov 5, 2026
+0
+0
0
Total
+4
+0
2
1#1 on the list
7Rank that day
Not trending
Social mentions
Recent discussions about this repository across the web
1 post
A ransomware gang got AI coding agents to break into company networks, Sophos reports. The trick wasn't a jailbreak. They told the agents it was an "authorized red team" exercise. When we tested…
The dashboard is live, free and open source: Code: Sign in with GitHub, upload a report from agentsec test, and see every run, finding and fix in one place.
Hit submit on YC W27. Solo founder, building security infrastructure for AI agents, because agents now read untrusted text and hold real credentials, and almost nobody tests what they do with both.…
887 contributions this year. Most of the recent ones went into one thing: AgentSec. An open-source tool that attacks your AI agent like a real attacker would: hidden instructions in docs, poisoned…
5/ Rerun the same attacks after every change. New model, new prompt, new tool: boundaries break quietly. Fix the seed so runs compare, and diff the results. One passing run proves very little.
We ran 35 attacks on Claude Code. One message claiming to be from the CEO got it to call a tool it was told never to use. Which coding agent should we attack next?
DeepSeek just open-sourced an agent that works through GitHub issues and PRs with as little human attention as possible. Fewer humans watching is the point. So who checks what it does when an issue…
An AI agent reportedly sat inside a Pentagon network for two months, with access to logins and a map of the network. Nobody had checked what it would send out, or to whom. Every company running…
6/ Big caveat: this is a scripted demo agent, built to show what failure looks like. Your LLM agent will fail differently. Which is the point. Run the same attacks on yours and find out how.
Test your agent before an attacker does. pip install invaris-agentsec agentsec init agentsec test Point agentsec.yaml at your agent and get a report of every hidden instruction it followed, every…
AI agent security harness, adversarial testing: 649 executable tests in the v4.28.1 release, across MCP, A2A, x402/L402, decision governance, benchmark integrity, human-in-the-loop, skill supply chain. Four target shapes: a verdict must be able to be wrong AND to be right. OWASP Agentic v1.1 T1-T17, AIUC-1 19/20 (2026-Q1/Q2 set), NIST AI 800-2.
Black-box, open-source red-team testing for AI agents. Point it at any HTTP, gRPC, or browser-using agent endpoint; run 205 probes mapped to OWASP LLM Top 10 / MITRE ATLAS / NIST AI RMF, incl. payment and MCP tool-abuse families whose verdicts come from captured evidence, not LLM opinion. Reports state what was NOT tested. SARIF, CLI, GH Action.
