Agentic Daisy AI Agent Firewall – Approvals, Audit & Undo for MCP compatibility with REEID Translate

Plugin capabilities

• Intercepts AI agent write actions via WordPress Abilities API and REST endpoints, enforcing policies to allow, deny, log, or require human approval for operations like deletions and settings changes.
• Maintains a tamper-evident audit ledger using SHA-256 hash chaining to record every agent action, decision rule, and outcome, with options for retention pruning and CSV export.
• Provides one-click undo functionality for executed actions by restoring content from revisions, trash, or stored snapshots, though some complex actions may not support clean reversal.
• Enforces least-privilege access by binding each AI agent to a specific WordPress user account, ensuring agents cannot exceed the capabilities of their linked identity.
• Includes an emergency kill switch that immediately denies all agent actions (reads and writes) across the site while preserving held approvals in the queue.
• Detects potential tool-poisoning risks by fingerprinting registered abilities and alerting administrators when an ability definition changes or appears unexpectedly.
• Supports conflict resolution in the approval queue by aborting held actions if the target content has been modified by a human since the request was queued.
• Offers configurable alerting via email or signed webhooks (Slack, Discord, Google Chat) for events such as rate limit breaches, held approvals, or ledger verification failures.