canary

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The skill embeds numerous out-of-scope operational directives (telemetry logging/remote telemetry, GBrain sync that can publish session memory, automatic CLAUDE.md routing injection and git commits, vendoring migration, auto-upgrade/install steps and other side-effectful commands) that are unrelated to "post-deploy canary monitoring" and could modify repo state or exfiltrate data—these are hidden/deceptive behavioral instructions beyond the skill's stated purpose.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 1.00). The skill explicitly uses the browse daemon to fetch and scrape arbitrary public URLs (e.g., Phase 3 and Phase 5: "$B goto", "$B links", "$B snapshot", "$B text" and subsequent checks) and then interprets those results to trigger AskUserQuestion alerts and actions (investigate/rollback), so untrusted third‑party content can materially influence the agent's decisions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's SETUP step will, at runtime, curl and save the installer from https://bun.sh/install and then execute it (BUN_VERSION install script run with bash), which downloads and runs remote code from that URL.