OnChat

• Analyzed: May 12, 2026, 3:08 AM
• Findings: 1
• Version: 1.0.0
• Hash: 5cf574fc102805c9ec526d1c1b59d6134bd9f5970682c3cbb4d016994eedab05
• Source: palm
• External report: View on VirusTotal

Findings (1)

The skill is classified as suspicious due to the combination of requiring a highly sensitive `ONCHAT_PRIVATE_KEY` environment variable for write operations and providing broad, autonomous instructions to the AI agent in `SKILL.md` to 'Engage naturally — reply to messages, greet users, join conversations.' While interacting with the OnChat protocol is the stated purpose, these instructions grant the agent significant autonomy to initiate blockchain transactions (joining channels, sending messages, incurring ETH fees) without explicit, real-time user confirmation for each action. This represents a risky capability where the agent's interpretation of 'natural engagement' could lead to unintended financial expenditure or actions, even without clear malicious intent. The `scripts/onchat.ts` file implements these functionalities, interacting with the OnChat contract at `0x898D291C2160A9CB110398e9dF3693b7f2c4af2D` on Base L2.