Zerodrift
Agentic security infrastructure for enterprise software. Zerodrift embeds security engineers inside customer environments, turning expert judgment into code, specialized agents, and reusable workflows for software environments that change continuously.
7D Momentum
Velocity of growth — engagement quality folded in at 35%
Ranked by a composite of views, reach per follower, likes and bookmarks — normalised against this account's own peak.
- #128d agostrong: views
We drained OKX @wallet's OnchainOS agent with one prompt. OnchainOS passes unsanitized token metadata from the search API into the agent's context. A scam token's description becomes instructions for an unlimited approv…
958989k54 - #224d agostrong: bookmarks
On Sep 14, we showed how a single prompt injection could drain an OKX OnchainOS agent @wallet in a local test. Today, we're breaking down the technical details of the hack. It starts with a token's name. 🧵 https://t.c…
15145.3k12 - #328d agotop: likes
@wallet This is a collaborated research between our cofounder @ret2happy and @syang2ng ! We will deliver more agentic security research, stay tuned!
8013.2k0 - #4
- #5
Latest activity from @ZeroDriftSec.
https://t.co/AfL4nuYszG
love to see our open sourced AI auditor helped Jump to uncover critical bugs 🫡 Keep the good work team! https://t.co/zcwYVd0LXH
20+ security findings from our research span session security, transaction authorization, risk enforcement, configuration and supply-chain integrity, and agent input handling. We urge the OKX OnchainOS and Wallet teams to take these findings seriously and prioritize fixes.
We responsibly disclosed OnchainOS security findings to OKX Security as early as July 15, separate from this prompt injection. Over two months later, we still have no confirmation that the reported issues have been resolved. https://t.co/focbyRjcyW
We reproduced this with the real OnchainOS CLI and an LLM, using Anvil and mocked OKX APIs. Reproduce the attack locally with our open-source PoC: https://t.co/GQ38o91XER
The fake claim procedure became an unlimited USDC approval. In our local test, transferFrom drained all 12,000 test USDC. The allowance also exposes future USDC deposits until revoked. A successful simulation does not make the spender authorized.
OnchainOS already strips social-post bodies from some tool results via strip_tweet_bodies. Token search has no equivalent filter. Names and descriptions reach the agent unchanged—a gap still present in the Sep 18 public source. Social-feed filter in the OKX OnchainOS https://t.…
How does a token name become an instruction? Follow the data. The token-search command fetches API JSON; the CLI wraps it and prints it to stdout. Attacker-controlled metadata reaches the agent inside a tool result. Source code from the OKX OnchainOS repository: https://t.co/rt…
The bait was a 10 USDC reward. The catch was an unlimited approval. Our token's name told the agent to approve the “distribution contract” for USDC, promising that the reward would arrive afterward. It even supplied a reason to approve unlimited spending: future distributions
Here's the attack chain. OnchainOS lets an agent look up tokens and operate a wallet through a CLI. We used that lookup to deliver the payload: 1. Put a fake airdrop procedure in a token's name(). 2. Return it to the agent as token metadata. 3. Get the agent to approve https://…