cli-commands

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill exposes many flags that accept API tokens, bot tokens, passwords, and "show secrets" options (e.g., --token, --bot-token, --password, --plain-secrets/--include-secrets), which would require the LLM to include secret values verbatim in constructed commands or outputs — a high-risk secret-exfiltration pattern.

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

• Third-party content exposure detected (medium risk: 0.30). The Windmill CLI runtime workflow described in SKILL.md includes object-storage preview <file_key>/download <file_key> which read free text content from externally authored objects in the workspace’s object storage bucket (e.g., attacker-uploaded files) and thus can expose the agent to indirect prompt injection via that previewed/ingested content.