deploying-honeytokens-and-canarytokens

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The workflow explicitly instructs fetching and executing remote code via "git clone https://github.com/thinkst/canarytokens-docker" followed by "docker compose up -d" (which runs the fetched containers), and the agent also performs runtime API calls to https://canarytokens.org for token generation; the GitHub clone + docker compose is a high-confidence execution-of-remote-code dependency.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs creating an Active Directory account (New-ADUser) and planting decoy credentials in system locations (e.g., ~/.aws/credentials, /etc/hosts) and deploying network services on privileged ports via Docker—actions that modify host/authentication state and create new user accounts.