AI • Agents • Open Source tools

Joined October 2023
Pinned Tweet
THIS IS F**KING DANGEROUS SOME COMPANIES ARE LEAVING WAY TOO MUCH EXPOSED I found 3 OSINT tools that can uncover a ridiculous amount of information from the outside. LEAKIX -> Give it a domain and it can surface publicly exposed services and databases connected to that infrastructure. ARGUS -> Runs 135+ recon and security checks against a website to build a much broader picture of its external footprint. GITGRABER -> Monitors GitHub for accidentally exposed secrets like API keys, passwords, and sensitive information —> with Telegram alerts when something is found. Different tools. Same idea: find what shouldn’t be publicly visible. If you’re learning OSINT, bug bounty, or pentesting, these are worth knowing. BOOKMARK it before someone take it down LINK’S BELOW
41
295
2
2,626
251,737
MAX retweeted
YOUR PHP SERVER COULD BE AT RISK OF REMOTE CODE EXECUTION Two critical PHP vulnerabilities. CVSS 9.2 each. This PHP2Shell repo brings together exploit research and defensive detection tools. → 5 documented RCE chains → PoCs for 3 exploit chains included → 13 Suricata/Snort detection rules → Sigma rules for SIEM monitoring → ModSecurity and OpenResty defenses → Docker lab for testing vulnerable configurations → Technical analysis and mitigation guidance The research targets PHP’s session.upload_progress handling and Zend Engine memory state issues, with some attack paths requiring no application vulnerability. The project reports affected versions from PHP 5.4.0 through 8.5.11. Patch to a fixed PHP release. Detection rules and configuration changes are not substitutes for patching. BOOKMARK IT & Only test systems you own or have explicit authorization to assess. REPO BELOW
21
10
2
88
5,361
YOUR PHP SERVER COULD BE AT RISK OF REMOTE CODE EXECUTION Two critical PHP vulnerabilities. CVSS 9.2 each. This PHP2Shell repo brings together exploit research and defensive detection tools. → 5 documented RCE chains → PoCs for 3 exploit chains included → 13 Suricata/Snort detection rules → Sigma rules for SIEM monitoring → ModSecurity and OpenResty defenses → Docker lab for testing vulnerable configurations → Technical analysis and mitigation guidance The research targets PHP’s session.upload_progress handling and Zend Engine memory state issues, with some attack paths requiring no application vulnerability. The project reports affected versions from PHP 5.4.0 through 8.5.11. Patch to a fixed PHP release. Detection rules and configuration changes are not substitutes for patching. BOOKMARK IT & Only test systems you own or have explicit authorization to assess. REPO BELOW
21
10
2
88
5,361
THIS IS F**KING DANGEROUS YOUR AI’S GUARDRAILS AREN’T AS STRONG AS YOU THINK. Basilisk is an open source AI red teaming framework built to expose weaknesses in LLMs. → 33 attack modules mapped to OWASP LLM Top 10 → evolves jailbreaks when prompts get refused → tests prompt injection, data leaks, tool abuse & RAG poisoning → fingerprints models and discovers available tools → compares GPT, Claude & Gemini side by side → grades guardrails from A+ to F → generates evidence backed findings and signed audit logs It can even run repeatable security tests to catch regressions. Only test systems you own or have explicit authorization to assess. REPO BELOW
39
31
1
308
15,763
MAX retweeted
THIS IS F**KING DANGEROUS YOUR AI’S GUARDRAILS AREN’T AS STRONG AS YOU THINK. Basilisk is an open source AI red teaming framework built to expose weaknesses in LLMs. → 33 attack modules mapped to OWASP LLM Top 10 → evolves jailbreaks when prompts get refused → tests prompt injection, data leaks, tool abuse & RAG poisoning → fingerprints models and discovers available tools → compares GPT, Claude & Gemini side by side → grades guardrails from A+ to F → generates evidence backed findings and signed audit logs It can even run repeatable security tests to catch regressions. Only test systems you own or have explicit authorization to assess. REPO BELOW
39
31
1
308
15,763
MAX retweeted
I’M BEGGING YOU TO TRY THIS WITH GROK BOT BEFORE BED. let it research the market, uncover strategy ideas, and build a trading bot while you sleep. wake up with something you can backtest, refine, and put to the test. BOOKMARK this prompt before it starts making the rounds
22
7
1
58
3,096
I’M BEGGING YOU TO TRY THIS WITH GROK BOT BEFORE BED. let it research the market, uncover strategy ideas, and build a trading bot while you sleep. wake up with something you can backtest, refine, and put to the test. BOOKMARK this prompt before it starts making the rounds
22
7
1
58
3,096
HACKERS DON’T WANT YOU TO SEE THIS FOUND 3 WIFI HACKING TOOLS THAT FEEL ILLEGAL TO KNOW ABOUT. → FragAttacks: exposes flaws in Wi-Fi protocols that can affect devices across generations. → AngryOxide: a stateful wireless attack engine that can adapt its approach during testing. → EAPHammer: simulates rogue enterprise Wi-Fi access points to test how easily credentials could be exposed. These tools show how attackers can exploit weaknesses in wireless security. Only test networks you own or have explicit authorization to assess. REPO BELOW
22
50
354
12,517
MAX retweeted
HACKERS DON’T WANT YOU TO SEE THIS FOUND 3 WIFI HACKING TOOLS THAT FEEL ILLEGAL TO KNOW ABOUT. → FragAttacks: exposes flaws in Wi-Fi protocols that can affect devices across generations. → AngryOxide: a stateful wireless attack engine that can adapt its approach during testing. → EAPHammer: simulates rogue enterprise Wi-Fi access points to test how easily credentials could be exposed. These tools show how attackers can exploit weaknesses in wireless security. Only test networks you own or have explicit authorization to assess. REPO BELOW
22
50
354
12,517
MAX retweeted
THIS IS ACTUALLY HUGE Drex 2.0 now runs completely local on Apple Silicon. No cloud, no API calls, no data leaving your machine. Here's what makes it wild: • Open weights, so you can run it, inspect it, and build on it yourself • Sub 200ms latency, fully offline • 21 GB footprint at 4-bit (Q4_K_M) • Runs on any Apple Silicon Mac with 36 GB+ unified memory And the speed numbers on an M5 Pro: • Kapitsa: 113 ms median decision latency • llama.cpp: 248 ms That's more than 2x faster than llama.cpp, on a laptop. Local AI used to mean slow, watered down models. Now it means private, fast, and zero cloud bill. Your MacBook just became the server
Drex 2.0 now runs completely local on Apple Silicon. Open weights. Sub-200ms latency. Zero cloud dependencies. • Footprint: 21 GB (4-bit Q4_K_M) • Hardware: 36 GB+ Unified Memory (Apple Silicon) Median decision latency (M5 Pro): • Kapitsa: 113 ms • llama.cpp: 248 ms • Ollama: 282 ms Hugging Face: huggingface.co/nace-ai/drex-… GitHub: github.com/nace-ai/drex-deci…
17
5
77
7,893
THIS IS ACTUALLY HUGE Drex 2.0 now runs completely local on Apple Silicon. No cloud, no API calls, no data leaving your machine. Here's what makes it wild: • Open weights, so you can run it, inspect it, and build on it yourself • Sub 200ms latency, fully offline • 21 GB footprint at 4-bit (Q4_K_M) • Runs on any Apple Silicon Mac with 36 GB+ unified memory And the speed numbers on an M5 Pro: • Kapitsa: 113 ms median decision latency • llama.cpp: 248 ms That's more than 2x faster than llama.cpp, on a laptop. Local AI used to mean slow, watered down models. Now it means private, fast, and zero cloud bill. Your MacBook just became the server
Drex 2.0 now runs completely local on Apple Silicon. Open weights. Sub-200ms latency. Zero cloud dependencies. • Footprint: 21 GB (4-bit Q4_K_M) • Hardware: 36 GB+ Unified Memory (Apple Silicon) Median decision latency (M5 Pro): • Kapitsa: 113 ms • llama.cpp: 248 ms • Ollama: 282 ms Hugging Face: huggingface.co/nace-ai/drex-… GitHub: github.com/nace-ai/drex-deci…
17
5
77
7,893
MAX retweeted
asked for an office and tavus built a place where nobody would ever get any work done HR could never
I asked Claude to build an office for the @tavus PALs... I was not expecting it to build THAT kind of office
13
2
56
4,014
asked for an office and tavus built a place where nobody would ever get any work done HR could never
I asked Claude to build an office for the @tavus PALs... I was not expecting it to build THAT kind of office
13
2
56
4,014
THIS IS F**KING INSANE I FOUND AN AI PENTESTING AGENT THAT DOESN’T JUST FIND VULNERABILITIES. IT TRIES TO EXPLOIT THEM. → analyzes source code and maps attack paths → tests for SQL injection, XSS, SSRF, broken authentication & authorization → uses browser automation + CLI tools to run real exploits → validates findings with proof-of-concept evidence → generates PDF, Markdown, JSON & SARIF reports → integrates with GitHub Actions and GitLab CI/CD → supports OpenAI, Anthropic, xAI, AWS Bedrock & local models → runs in your own infrastructure with BYOK only test systems you own or have explicit written authorization to assess. Repo below
33
67
4
436
23,381
MAX retweeted
THIS IS F**KING INSANE I FOUND AN AI PENTESTING AGENT THAT DOESN’T JUST FIND VULNERABILITIES. IT TRIES TO EXPLOIT THEM. → analyzes source code and maps attack paths → tests for SQL injection, XSS, SSRF, broken authentication & authorization → uses browser automation + CLI tools to run real exploits → validates findings with proof-of-concept evidence → generates PDF, Markdown, JSON & SARIF reports → integrates with GitHub Actions and GitLab CI/CD → supports OpenAI, Anthropic, xAI, AWS Bedrock & local models → runs in your own infrastructure with BYOK only test systems you own or have explicit written authorization to assess. Repo below
33
67
4
436
23,381