d6e-plugin-development

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The documents describe several platform capabilities that can be straightforwardly abused for intentional data exfiltration, credential collection, remote callbacks, and supply‑chain/backdoor updates (unbounded effect HTTP, install-time secret prompts, and installer/runtime behavior that follows latest versions), so plugin manifests and installs are a high-risk attack surface if not strictly reviewed and constrained.

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

• Third-party content exposure detected (medium risk: 0.30). The skill documentation describes plugin “Fetch” input steps that can retrieve and ingest external JSON from attacker-influenced URLs at workflow execution time (via workflows[].input_steps[].source.type: Fetch → external HTTP response parsed and provided to later steps/STFs).

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

• Potentially malicious external URL detected (high risk: 1.00). The doc states that d6e instances fetch the marketplace registry (https://marketplace.d6e.ai/api/registry) and the manifestUrl/raw template (https://raw.githubusercontent.com/your-org/d6e-plugin-your-plugin/v1.0.0/template.yaml) at install/runtime, and those remote template.yaml files include template_prompt which is injected into the AI agent's system prompt — meaning fetched external content can directly control agent prompts.