Four tools with complete schemas and descriptions, but descriptions are domain-specific and lack LLM-optimized guidance. All tools are READ_ONLY (good security posture). Schemas are well-formed with typed parameters and defaults. However, descriptions focus on what the API returns rather than when/why an LLM should call the tool. No error handling guidance, no parameter constraints (enums), and no output schema documentation. Tool names are clear (verb_noun pattern) but descriptions average ~150 chars and lack actionable context for agent selection.

• Expand descriptions to 100-200 chars and add explicit WHEN guidance: 'Call this to retrieve the agent's current identity and lifecycle state before checking permissions.' vs. 'Call this to understand why a specific decision was allowed or denied.'
• Document output schemas for each tool. E.g., whoami returns {agent_id, tenant, definition, lifecycle_state, identity_assurance, trust_domain}; authority returns {declared, granted, observed, undeclared, ungranted, capability_outside_authority, unused_authority}.
• Add error recovery hints to descriptions: 'If agent_id is not found, use the agent registry API to list available agents.' or 'If decision_id is invalid, check the audit log for recent decisions.'
• Add a description to the 'tenant' parameter explaining multi-tenancy: 'The tenant identifier (default: "default"). Override if querying a different tenant's agent.'
• Consider adding a discovery tool (e.g., list_agents or search_agents) to help LLMs find valid agent_ids without trial-and-error.