PriorMCP has 10 tools with complete input schemas and descriptions, but quality is uneven. All tools follow verb_noun naming (verify_coverage, check_prior_auth, etc.) and include parameter descriptions. However, descriptions are generic (avg ~60 chars, well below the 194-char baseline for A+ tools) and lack actionable context about when to use each tool vs. alternatives. No output schemas are documented, LLMs cannot predict response structure for downstream chaining. Patient lookup tools (GetPatientAge, GetPatientAllergies, FindPatientId) accept optional patientId but lack guidance on resolution order or error recovery. Coverage/auth tools (verifyCoverage, checkPriorAuth, estimateCost) have overlapping responsibilities without clear composition guidance. No error handling patterns visible, tools return strings but no structured error classification or recovery hints.

• Add output schemas to all tools. Document the structure of returned text (e.g., verifyCoverage returns {coverage: string, prior_auth_required: boolean, estimated_patient_cost: string, reason: string}). This enables LLMs to extract fields and chain tools.
• Expand tool descriptions to 100-200 chars. Include WHEN to use each tool and how it differs from similar ones. E.g., 'Checks if a treatment is covered under a plan. Use this before checkPriorAuth to determine coverage status. Returns coverage decision and estimated cost.'
• Add error handling guidance. Return structured errors with classification (retryable, user-fixable, fatal) and recovery hints. E.g., 'Patient not found. Try FindPatientId(firstName, lastName) first.' or 'Invalid plan name. Available plans: [list].'
• Document patient context resolution. Clarify that GetPatientAge and GetPatientAllergies can use FHIR context if patientId is omitted, and what happens if both are missing. Add a dependency hint: 'If you only have a name, call FindPatientId first.'
• Clarify tool composition. Add a note to assessTreatmentAccess explaining it runs the full workflow (coverage → prior auth → cost → alternatives → draft). Recommend using individual tools for targeted queries and assessTreatmentAccess for comprehensive assessment.