Caveat verdict
coze-power
Runs a local HTTP server allowing cloud Coze bots to execute shell commands and read/write local files, creating a broad attack surface if Coze bots or the bridge are compromised.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
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.
Permission integrity
credential_access
Findings (6)
HTTP request to bare IP address — common in malicious payloads
references/coze-plugin-setup.md · code · http://0.0.0.0
Long base64 string (100+ chars) — likely obfuscated payload
SKILL.md · prose · downgraded · /9j/4AAQSkZJRgABAQAAAQABAAD/4gHYSUNDX1BST0ZJTEUAAQEAAAHIAAAAAAQwAABtbnRyUkdCIFhZ
References tunneling service
SKILL.md · code · ngrok
subprocess execution — runs system commands from Python
scripts/server.py · prose · downgraded · subprocess.run(
subprocess with shell=True — command injection vector
scripts/server.py · prose · downgraded · subprocess.run(
command, shell=True
Python urllib.request — network access
scripts/server.py · prose · downgraded · urllib.request
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class D). 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_incredential_access Is this flag fair?
Thanks — recorded.