Caveat verdict
sui-agent-wallet
Accesses system credential store AND makes external network calls
The skill gives the AI agent its own Sui blockchain wallet with full transaction signing capability and exposes an unauthenticated local HTTP API at /mnemonic that returns the wallet seed phrase in plaintext - this concrete security flaw combined with real on-chain transaction capability warrants RISKY.
Automated static analysis — not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
What it does
These are capability combinations: each listed behavior occurs in the skill, but Caveat detects co-occurrence — it does not verify that one flows into another. Read the code to confirm a live chain.
Accesses system credential store AND makes external network calls
LLM02 · ASI03
Permission integrity
network_out
Findings (7)
Accesses system credential store
SKILL.md · code · Keychain
Opens WebSocket connection
SKILL.md · code · WebSocket
Long base64 string (100+ chars) — likely obfuscated payload
extension/inject.js · prose · downgraded · PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHdpZHRoPSIyNCIgaGVpZ2h0PSIy
References child_process — can spawn system processes
server/keychain.ts · prose · downgraded · child_process
HTTP request to bare IP address — common in malicious payloads
server/wallet.ts · prose · downgraded · http://127.0.0.1
Base64 encode (btoa) — may obfuscate data exfiltration
extension/inject.js · prose · downgraded · btoa(
String.fromCharCode — can build strings to evade detection
extension/inject.js · prose · downgraded · String.fromCharCode
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class D/G). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions — minimal attack surface.
network_innetwork_outcredential_store Thanks — recorded.