addedOutput schema / $defs
Added value: +{
+ "ErrorDetail": {
+ "description": "Structured failure body. Codes mirror app/exceptions.AppException\nsubclasses; agent retry / upgrade decisions key off `code`, not `message`.",
+ "properties": {
+ "code": {
+ "description": "Stable machine-readable failure category. Agents key retry/upgrade decisions off this.",
+ "enum": [
+ "invalid_argument",
+ "not_found",
+ "rate_limit_exceeded",
+ "auth_required",
+ "tier_limit",
+ "upstream_timeout",
+ "upstream_error",
+ "internal_error"
+ ],
+ "title": "Code",
+ "type": "string"
+ },
+ "docs_url": {
+ "anyOf": [
+ {
+ "type": "string"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "Documentation pointer (e.g. tool input contract) when code='invalid_argument'.",
+ "title": "Docs Url"
+ },
+ "message": {
+ "description": "Human-readable detail. Free text — never parse. Capped at 500 chars to prevent oversized upstream errors from bloating responses.",
+ "maxLength": 500,
+ "title": "Message",
+ "type": "string"
+ },
+ "retry_after_seconds": {
+ "anyOf": [
+ {
+ "type": "integer"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "When code='rate_limit_exceeded', the minimum seconds to wait before retrying.",
+ "title": "Retry After Seconds"
+ },
+ "upgrade_url": {
+ "anyOf": [
+ {
+ "type": "string"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "Pricing/upgrade URL when code='tier_limit' or 'rate_limit_exceeded' on the Free tier.",
+ "title": "Upgrade Url"
+ }
+ },
+ "required": [
+ "code",
+ "message"
+ ],
+ "title": "ErrorDetail",
+ "type": "object"
+ },
+ "ErrorResponse": {
+ "description": "MCP error envelope. Tool return type is always\n`SpecificResponse | ErrorResponse` — Union flag tells the agent which arm\narrived without parsing the inner body.",
+ "properties": {
+ "error": {
+ "$ref": "#/$defs/ErrorDetail"
+ }
+ },
+ "required": [
+ "error"
+ ],
+ "title": "ErrorResponse",
+ "type": "object"
+ },
+ "PivotHint": {
+ "additionalProperties": true,
+ "description": "A suggested follow-up MCP tool call. Surfaced inside response.next_calls so\nLLM agents can chain related lookups without manual prompting. Each hint names\nthe tool, the input value to pass, and a short reason explaining why this\npivot adds value in the current context.",
+ "properties": {
+ "input": {
+ "description": "Suggested input value to pass to the tool — typically a CVE ID, CWE ID, domain, or IP. Pre-populated from the current response so the agent can call the next tool without re-deriving the argument.",
+ "title": "Input",
+ "type": "string"
+ },
+ "params": {
+ "anyOf": [
+ {
+ "additionalProperties": {
+ "type": "string"
+ },
+ "type": "object"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "Optional extra kwargs to pass alongside `input`. Used by pivot generators when the next call benefits from a secondary parameter, e.g. {'exclude_id': 'AML.T0051'} to skip the originating technique from a sibling-tactic search. Omitted when no extra args are needed.",
+ "title": "Params"
+ },
+ "reason": {
+ "description": "Short rationale (one sentence) for why this follow-up call adds value, e.g. 'Federal patch deadline + ransomware association', 'Public exploits / PoC availability'.",
+ "title": "Reason",
+ "type": "string"
+ },
+ "tool": {
+ "description": "Canonical MCP tool name to call next. Constrained to known operation_ids in tools/list — adding a new tool here requires expanding the Literal.",
+ "enum": [
+ "cve_lookup",
+ "cve_search",
+ "cve_leading",
+ "bulk_cve_lookup",
+ "exploit_lookup",
+ "kev_detail",
+ "cwe_lookup",
+ "subdomain_enum",
+ "ssl_check",
+ "tech_fingerprint",
+ "asn_lookup",
+ "ip_lookup",
+ "ioc_lookup",
+ "bulk_ioc_lookup",
+ "hash_lookup",
+ "threat_intel",
+ "threat_report",
+ "audit_domain",
+ "domain_report",
+ "dns_lookup",
+ "whois_lookup",
+ "wayback_lookup",
+ "scan_headers",
+ "check_headers",
+ "check_secrets",
+ "check_injection",
+ "check_dependencies",
+ "email_mx",
+ "email_disposable",
+ "phone_lookup",
+ "username_lookup",
+ "password_check",
+ "phishing_check",
+ "atlas_technique_lookup",
+ "atlas_technique_search",
+ "bulk_atlas_technique_lookup",
+ "atlas_case_study_lookup",
+ "atlas_case_study_search",
+ "d3fend_defense_lookup",
+ "d3fend_defense_search",
+ "d3fend_defense_for_attack",
+ "d3fend_attack_coverage"
+ ],
+ "title": "Tool",
+ "type": "string"
+ }
+ },
+ "required": [
+ "tool",
+ "input",
+ "reason"
+ ],
+ "title": "PivotHint",
+ "type": "object"
+ },
+ "UsernameLookupResponse": {
+ "properties": {
+ "checked_count": {
+ "default": 0,
+ "description": "Number of platforms actually checked (may be less than total platforms if early-exit).",
+ "title": "Checked Count",
+ "type": "integer"
+ },
+ "error": {
+ "anyOf": [
+ {
+ "type": "string"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "Input validation error (empty username, invalid chars, too long). Null on successful lookups.",
+ "title": "Error"
+ },
+ "found_count": {
+ "default": 0,
+ "description": "Number of platforms where status=='found'.",
+ "title": "Found Count",
+ "type": "integer"
+ },
+ "next_calls": {
+ "anyOf": [
+ {
+ "items": {
+ "$ref": "#/$defs/PivotHint"
+ },
+ "type": "array"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "Suggested follow-up MCP tool calls. Ordered by relevance; agents should chain these without re-prompting the user.",
+ "title": "Next Calls"
+ },
+ "results": {
+ "description": "Per-platform results, sorted: found first, then alphabetical by platform.",
+ "items": {
+ "$ref": "#/$defs/UsernameMatch"
+ },
+ "title": "Results",
+ "type": "array"
+ },
+ "summary": {
+ "default": "",
+ "description": "One-line human-readable summary, e.g. 'username \"x\" found on 6/20 platforms (3 unavailable)'.",
+ "title": "Summary",
+ "type": "string"
+ },
+ "username": {
+ "default": "",
+ "description": "Echoed normalized username (lowercased, validated against [a-z0-9._-]).",
+ "title": "Username",
+ "type": "string"
+ },
+ "verdict": {
+ "anyOf": [
+ {
+ "$ref": "#/$defs/Verdict"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "Falsifiability metadata: sources_queried, sources_unavailable, completeness, deterministic flag. Lets agents distinguish 'no data' from 'source failed' without re-running the call."
+ }
+ },
+ "title": "UsernameLookupResponse",
+ "type": "object"
+ },
+ "UsernameMatch": {
+ "properties": {
+ "platform": {
+ "default": "",
+ "description": "Platform identifier, e.g. 'github', 'twitter', 'reddit'.",
+ "title": "Platform",
+ "type": "string"
+ },
+ "status": {
+ "default": "error",
+ "description": "Per-platform outcome. 'found'/'not_found' are terminal factual answers. 'rate_limited' (429), 'blocked' (403 — often Cloudflare/bot detection), 'timeout' (network), and 'error' (5xx/other) are unavailability states — the platform's answer is unknown, NOT 'user does not exist'. Agents should treat these four as sources_unavailable, not negative evidence.",
+ "enum": [
+ "found",
+ "not_found",
+ "rate_limited",
+ "blocked",
+ "timeout",
+ "error"
+ ],
+ "title": "Status",
+ "type": "string"
+ },
+ "url": {
+ "default": "",
+ "description": "Canonical profile URL for this platform+username (may 200/redirect even when not_found).",
+ "title": "Url",
+ "type": "string"
+ }
+ },
+ "title": "UsernameMatch",
+ "type": "object"
+ },
+ "Verdict": {
+ "properties": {
+ "completeness": {
+ "default": "complete",
+ "description": "'complete' = every planned source returned data; 'partial' = at least one source in sources_unavailable failed or was skipped; 'minimal' = only the primary/required source returned, optional enrichment missing.",
+ "enum": [
+ "complete",
+ "partial",
+ "minimal"
+ ],
+ "title": "Completeness",
+ "type": "string"
+ },
+ "data_age_seconds": {
+ "anyOf": [
+ {
+ "type": "integer"
+ },
+ {
+ "type": "null"
+ }
+ ],
+ "default": null,
+ "description": "Seconds elapsed since the oldest cached source was fetched, or null when every source was queried live for this request. Use to judge freshness.",
+ "title": "Data Age Seconds"
+ },
+ "deterministic": {
+ "description": "True when the response is fully reproducible from the listed sources for the same input at the same moment (no randomness, no model inference). False for endpoints that include probabilistic scoring or LLM output.",
+ "title": "Deterministic",
+ "type": "boolean"
+ },
+ "falsifiable_fields": {
+ "description": "Top-level response fields whose values a caller can independently re-derive from the named upstream sources (e.g. 'dns', 'ssl', 'whois'). Fields not in this list are derived/computed and cannot be directly re-verified.",
+ "items": {
+ "type": "string"
+ },
+ "title": "Falsifiable Fields",
+ "type": "array"
+ },
+ "sources_queried": {
+ "description": "Canonical source identifiers successfully consulted for this response (e.g. 'ripe_stat', 'shodan_internetdb', 'firehol'). Agent-readable list, order not significant.",
+ "items": {
+ "type": "string"
+ },
+ "title": "Sources Queried",
+ "type": "array"
+ },
+ "sources_unavailable": {
+ "description": "Sources that were expected but not returned — either intentionally skipped (lite mode, tier gating) or failed (quota, timeout, upstream down). Empty list means every planned source produced data.",
+ "items": {
+ "type": "string"
+ },
+ "title": "Sources Unavailable",
+ "type": "array"
+ }
+ },
+ "required": [
+ "deterministic"
+ ],
+ "title": "Verdict",
+ "type": "object"
+ }
+}
