{"ok":true,"site":{"host":"aidefend.net","url":"https://aidefend.net/","type":"live","desc":"AIDEFEND is a practical knowledge base mapping AI security defense techniques across 9 major frameworks including MITRE ATLAS, OWASP Top 10 for LLM, ML, and Agentic Applications, MAESTRO, NIST…","tools":[{"name":"search_techniques","kind":"answer","impl":"imperative","description":"Search the AIDEFEND AI security defense framework by keyword. Searches 367 taxonomy entities: 58 non-actionable parent families plus 309 actionable standalone/sub-technique controls across 7 tactics. Results identify whether an entity is a parent family or an actionable control. Framework content is in English; translate non-English queries before calling this tool.","annotations":{"readOnlyHint":true},"inputSchema":{"type":"object","properties":{"query":{"type":"string","description":"Search keyword in English, e.g. 'prompt injection', 'model scanning', 'adversarial'. IMPORTANT: The framework content is in English. If the user query is in another language (e.g., Traditional Chinese), MUST translate the keywords into English before calling this tool.","minLength":2,"maxLength":200},"maxResults":{"type":"integer","description":"Maximum results to return per category (techniques and sub-techniques). Default: 10, max: 30","default":10,"minimum":1,"maximum":30}},"required":["query"]},"hasHandler":true,"hasExecute":true,"__source":"spec","executable":true,"handlerField":"execute","page":"/","pages":["/"]},{"name":"get_technique_detail","kind":"answer","impl":"imperative","description":"Get full details of an AIDEFEND technique or sub-technique by its ID. Returns description, optional structured scope boundary, threat mappings across 9 security frameworks (MITRE ATLAS, MAESTRO, the OWASP Top 10 frameworks for LLM, ML, and Agentic Applications, NIST AML, Cisco, Google SAIF 2.0, and Databricks DASF 3.0), pillar/phase classification, and sub-technique list.","annotations":{"readOnlyHint":true},"inputSchema":{"type":"object","properties":{"id":{"type":"string","description":"Technique ID (e.g. 'AID-M-001') or sub-technique ID (e.g. 'AID-M-001.001')","pattern":"^AID-[A-Z]{1,2}-\\d{3}(\\.\\d{3})?$"}},"required":["id"]},"hasHandler":true,"hasExecute":true,"__source":"spec","executable":true,"handlerField":"execute","page":"/","pages":["/"]},{"name":"get_implementation_guide","kind":"answer","impl":"imperative","description":"Get practical implementation guidance for an AIDEFEND defense: step-by-step strategies with optional code examples, plus recommended OSI open-source, source-available/open-weight, and commercial/hosted tools. Accepts technique ID (e.g. AID-H-001) or sub-technique ID (e.g. AID-H-001.002). Use mode='full' for detailed code examples.","annotations":{"readOnlyHint":true},"inputSchema":{"type":"object","properties":{"id":{"type":"string","description":"Technique or sub-technique ID","pattern":"^AID-[A-Z]{1,2}-\\d{3}(\\.\\d{3})?$"},"mode":{"type":"string","description":"Output detail level. 'summary' returns strategy text only (shorter, fewer tokens). 'full' includes howTo details with code examples (longer but more actionable).","enum":["summary","full"],"default":"summary"}},"required":["id"]},"hasHandler":true,"hasExecute":true,"__source":"spec","executable":true,"handlerField":"execute","page":"/","pages":["/"]},{"name":"find_defenses_by_threat","kind":"answer","impl":"imperative","description":"Find AIDEFEND mappings for a specific threat. Returns actionable standalone/sub-technique controls separately from non-actionable parent-family context, so a family is never presented as an implementable control. Parent-family mappings are a derived union of reviewed child mappings for navigation, not independent control claims. Accepts exact or prefix framework IDs and English mapping-name keywords; exact IDs are preferred. OWASP LLM references default to the current edition, and explicit 2025 IDs or legacy names are migrated by risk concept rather than by rank. Malformed explicit editions and queries containing multiple OWASP LLM concepts return resolution metadata instead of being guessed. Cisco references resolve only against v2.0.0 with explicit edition metadata; legacy editions, retired IDs, multiple IDs and ID/name conflicts produce resolution metadata without a lookup. A mapping expresses a reviewed defensive relationship, not certification that a control is deployed or effective in the caller's environment.","annotations":{"readOnlyHint":true},"inputSchema":{"type":"object","properties":{"threat":{"type":"string","description":"Threat ID or keyword in English. Examples: 'AML.T0051', 'LLM01', 'LLM03:2025 Supply Chain', 'prompt injection', 'supply chain', 'model poisoning'. OWASP LLM 2025 references are accepted and resolved to their 2026 semantic successor. IMPORTANT: The framework content is in English. Translate any non-English queries into English before calling this tool.","minLength":2,"maxLength":200},"maxResults":{"type":"integer","description":"Maximum results per category. Default: 15, max: 30","default":15,"minimum":1,"maximum":30}},"required":["threat"]},"hasHandler":true,"hasExecute":true,"__source":"spec","executable":true,"handlerField":"execute","page":"/","pages":["/"]},{"name":"get_framework_stats","kind":"answer","impl":"imperative","description":"Population-explicit AIDEFEND statistics: taxonomy rows, top-level techniques, non-actionable parent families, standalone techniques, sub-techniques, actionable controls, mapping coverage for both all entities and actionable controls, leaf-only pillar/phase distributions, and canonical distinct-tool counts.","annotations":{"readOnlyHint":true},"inputSchema":{"type":"object","properties":{},"required":[]},"hasHandler":true,"hasExecute":true,"__source":"spec","executable":true,"handlerField":"execute","page":"/","pages":["/"]}],"category":"Developer Tools","favicon":"/favicons/aidefend.net.png","apiSurface":"spec","toolCount":5}}