Sallyport defines 3 tools with complete input schemas and descriptions. All tools have clear, security-focused descriptions (194-280 chars, within baseline 34-392). Schemas are properly structured with required fields and type constraints. However, output schemas are entirely undocumented, LLMs cannot plan downstream calls without knowing what fields to expect. Parameter descriptions are adequate but lack format/constraint details (e.g., 'timeout_s' has no min/max bounds). Error handling is mentioned in descriptions but not formalized. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles (http.request is READ_ONLY, ssh.exec is WRITE). Tool naming is strong (verb_noun pattern: http.request, ssh.exec, request_credential).

• Document output schemas for all three tools. For http.request, specify: {status: number, final_url: string, json?: object, body?: string}. For ssh.exec: {stdout: string, stderr: string, exit_code: number}. For request_credential: {provisioned: boolean, name?: string, message: string}.
• Add readOnlyHint and destructiveHint annotations to inputSchema. Mark http.request with readOnlyHint=true; mark ssh.exec with destructiveHint=true to signal agent safety constraints.
• Constrain timeout_s parameter: add minValue=1, maxValue=3600 to schema and note in description: 'Timeout in seconds (1-3600; default 30).'
• Formalize error handling: document expected error codes (e.g., 'host_not_found', 'auth_failed', 'timeout') and recovery guidance in tool descriptions.
• Clarify 'hosts' array format in request_credential: specify whether entries are domain names (api.example.com), FQDNs, or IPs; note that duplicates are deduplicated; add example: ['api.service.com', 'webhook.service.com'].