Darinor
darinor.com
Notes on building AI agents that don't leak, break, or turn into liabilities. https://darinor.com 🔗 AI + cyber starter pack: https://bsky.app/starter-pack/darinor.com/3ms6xkwmnlh2l
👤 Joined on 05 June 2026
121
Followers
360
Following
159
Total Posts
📜 Kullanıcı Geçmişi
- 05 Haz 2026, 08:17
🎉 Bluesky'e katıldı - İlk kullanıcı adı:
@syncdefend.bsky.social - 05 Haz 2026, 08:25
✏️ Kullanıcı adı değişti:
@syncdefend.bsky.social→@priamsec.bsky.social - 21 Tem 2026, 17:06
✏️ Kullanıcı adı değişti:
@priamsec.bsky.social→@darinorcom.bsky.social - 03 Ağu 2026, 16:48
✏️ Kullanıcı adı değişti:
@darinorcom.bsky.social→@darinor.com
🚫 Blokladığı Kullanıcılar
❤️ Beğendiği Gönderiler
Containment earns its keep when an agent is mistaken. A boundary that rejects unapproved credential use gives approval something concrete to lean on.
Show Post
The security question for agent tools is not just 'does it work?' It is 'what can it do when it is wrong, compromised, or over-permissioned?' #AIAgents #AppSec
Show Post
Seems it was a vibe-coded public app + prompt injection + an org where burning the equivalent of $600.000 in tokens is just your regular Tuesday. metr.org/blog/2026-08...
Show Post„prompted an agent directly to reveal it’s model provider API key“ Ok, maybe asking nicely shouldn’t count, depends on how broad you define prompt injection. In tend to go with confusing the agent into harmful action.
Show Post
It did. It submitted over 500 malicious packages. It's fantastic that Ruby gems stopped them. But it should never have gotten to that.
Show Post