Axeploit
Axeploit

Blog· 209 articles

Field notes from the offensive side

Attack chains, vulnerability deep dives, and hard-won lessons in API security from the Axeploit team.

"No Watermark Detected" Tells You Nothing. The SynthID Removal Toolchain Is Why
Latest12 min read

"No Watermark Detected" Tells You Nothing. The SynthID Removal Toolchain Is Why

By Jason Miller

Read article
Your Sent Items Folder Is a Phishing Kit Now
10 min read

Your Sent Items Folder Is a Phishing Kit Now

By Jason Miller

Grok Decrypted Its Own Attack Instructions. Your Agent Would Too
11 min read

Grok Decrypted Its Own Attack Instructions. Your Agent Would Too

By Jason Miller

The Hugging Face Sandbox Escape: Everyone Watched the Proxy, Nobody Watched Port 53
11 min read

The Hugging Face Sandbox Escape: Everyone Watched the Proxy, Nobody Watched Port 53

By Jason Miller

Five of the Top Seven Skills Were Malware: A Working Guide to OWASP's Agentic Skills Top 10
13 min read

Five of the Top Seven Skills Were Malware: A Working Guide to OWASP's Agentic Skills Top 10

By Jason Miller

CircleCI's MCP Server RCE: A Defender's Playbook for All Three Advisories
9 min read

CircleCI's MCP Server RCE: A Defender's Playbook for All Three Advisories

By Jason Miller

9,300 Leaked AWS Keys Are Still Live. One of Them Might Be Yours.
10 min read

9,300 Leaked AWS Keys Are Still Live. One of Them Might Be Yours.

A same-day runbook with copy-paste commands to find live leaked keys in your accounts and code, rotate them without downtime, and replace long-lived credentials with roles.…

By Jason Miller

RedC2 4.0 on npm: Detecting the Import-Time Linux Backdoor Behind 14 Trojanized Packages
11 min read

RedC2 4.0 on npm: Detecting the Import-Time Linux Backdoor Behind 14 Trojanized Packages

Get a concrete IoC set, ready-to-adapt Sigma and YARA rules, and a same-day audit and remediation workflow for the RedC2 4.0 npm campaign, plus a clear-eyed read on what its AI-driven C2 actually changes for detection.…

By Jason Miller

Self-Correction Loops Can Make LLM Pipelines Worse: An 85% to 62% Case Study and a Pre-Ship Measurement Playbook
13 min read

Self-Correction Loops Can Make LLM Pipelines Worse: An 85% to 62% Case Study and a Pre-Ship Measurement Playbook

A concrete failure taxonomy and a pre-ship measurement playbook (flip-rate tracking, shadow mode, paired holdout A/B, written ship gates) to decide whether a judge loop helps or quietly degrades their pipeline.…

By Jason Miller