orchestrator-approval

A verbatim approval workflow blocks subagent actions by presenting structured prompts with labeled fields like operation, command, scope, risk, and rollback. It ensures users cannot proceed without explicit consent and prevents decisions based on summaries or obfuscated data.

You can enforce verbatim command content by applying an orchestrator that intercepts mid-task approval requests. It presents structured prompts with labeled fields like operation, command, scope, risk, and rollback, ensuring users see the exact literal content before granting consent.

A nonce-based approval option in workflow governance is a unique identifier attached to each approval request. It enforces fresh approvals per blocked command by requiring users to select an option labeled 'Approve -- {action} [P-{nonce}]', preventing replay attacks or unauthorized batch approvals.

Yes, the orchestrator handles fresh approvals per blocked command and supports batch scenarios as needed. It enforces verbatim content in user approvals for subagent REVIEW flows, ensuring each action within a batch requires explicit consent through its own structured prompt.

No, you cannot use this approval workflow without frontmatter-driven metadata in SKILL.md. The orchestrator relies on this metadata to define invocability and discipline type, which is essential for enforcing the structured approval process and nonce-based options.

Subagent actions are blocked to force explicit user approvals with verbatim commands. This prevents decisions based on summaries or obfuscated data by ensuring the user is shown the literal content of any approval request and cannot proceed without explicit consent.