hivemind-goals 監査レポート.

警告

• Dependency or permission surface needs review
• Permission surface may require sandboxing
• The skill does not warn agents to treat goal/KPI file bodies and CLI output as untrusted data. A malicious or accidental instruction inside a stored Markdown file could be interpreted as a command by the agent.
• The documented `hivemind goal add --agent capture "..."` pattern relies on the agent correctly quoting user-supplied text; the skill should explicitly call out shell-injection-safe argument handling.
• Quality score needs review
• Permission surface needs review: secrets or environment access, shell or command execution
• Dependency/runtime risk: command execution surface, credential or environment access
• Permission surface: secrets or environment access, shell or command execution

方法

このレポートは公開メタデータ、AI レビュー、リポジトリの鮮度、インストール準備、OpenAgentSkill イベント、品質スコア、信頼チェック、Agent セーフティゲートを統合します。完全なソースコード監査ではありません。