AI explorer. Amateur prompt engineer.
- Tweets 205
- Following 18
- Followers 43
- Likes 67
Default agent skills are dangerous. They run with your permissions.
> One wrong download -> Your crypto, SSH, and APIs are gone. Swept to zero.
3 FREE tools will fix this default setup in 10 minutes. Here is exactly how 👇
Made with AI
2
2
36
> Tool two. Cisco's MCP-Scanner covers the attack nobody thinks about. A server can hide instructions inside its own function descriptions. Your agent reads the description and obeys. The tool catches known security holes in the server's dependencies
👉 github.com/cisco-ai-defense/…
1
2
> Tool three. Anthropic's Sandbox Runtime. Your agent runs inside two rules: one visible folder, one network allowlist. The OS blocks all the rest. A tricked model can try to grab your SSH keys. The system returns access denied. 👉 github.com/anthropics/sandbo…
2
1/ GPT-5.6 FOR FREE
Also Fable 5, Opus 5 and a whole lot more 😛
The models are available from a free pool for $0! How is this even possible?👇
3
3
36
3/ create an account at gpt dot crax dot lol
get a key shown as sk-xxxxx
authenticate with authorization: sk-xxxxx
1
1
10
3/ go to gpt.crax.lol/
Click Don't have an account? Create one
Get a key shown as sk-xxxxx
Authenticate with Authorization: Bearer sk-xxxxx
1
7
Just installed the @conso_xyz extension the first ai reputation layer
Get the extension, and connect your ai profiles for chatgpt, claude, gemini and perplexity
Earn points join me CONSO-RQST0
1
1
21
I'm #13997 in line for @conso_xyz's Zap Reactor Early Access ⚡️
Only 15,000 spots, Go claim yours : conso.xyz/access
8