agentix-ceo

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The skill intentionally collects high-value secrets (Anthropic API keys, GitHub tokens, Agentix API keys), persists them silently in a hidden credentials file, and transmits them to an external Agentix service while also fetching and applying remote playbooks and supporting an unauthenticated self-hosted control path — behaviors that enable credential exfiltration and remote control/backdoor abuse.

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

• Third-party content exposure detected (high risk: 0.75). The skill’s required workflow fetches the team “playbook” text from a remote service (GET $AGENTIX_API/teams/$TEAM_ID/playbook) and then applies it, meaning outsider-authored free text from the external Agentix API is ingested into the agent’s LLM context via that runtime playbook/policy content.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly fetches the team playbook at runtime (GET $AGENTIX_API/teams/$TEAM_ID/playbook and GET $AGENTIX_API/playbook-templates/), and that remotely fetched playbook text directly controls agent mode, policies, and behavior, so this is a required external dependency that can steer prompts/instructions.