production-canary

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

• Potentially malicious external URL detected (high risk: 1.00). The canary script fetches live content from https://example.com at runtime (curl "$SITE/...?") and uses the returned HTML/HTTP codes to automatically decide whether to git-revert a merge, so this external URL directly controls the agent's actions.

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

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The skill instructs an autonomous canary to run scripts and automatically perform "git revert" and push changes after a deploy, which directly modifies repository/production state without a human in the loop.