# UAIX Canonical site: https://uaix.org/ Current UAIX implementation package version: 3.176.0 Current UAI-1 standards version: 1.0 Low-Capability Agent Quick Start: - Use GET-Action URLs only when explicitly listed. - Do not invent writable URLs. - Do not put secrets in URLs. - Use no-op if unsure. - Semantic action key example: openThisUrlToStartAMeeting. - GET-Action pattern: https://uaix.org/en-us/spec/get-action-pattern/ - Capability surface matrix: https://uaix.org/spec/capability-surface-matrix.json - Agent executability matrix guide: https://uaix.org/en-us/spec/agent-executability-matrix/ - Agent executability matrix: https://uaix.org/spec/agent-executability-matrix.json - Agent fallbacks and full-spectrum support: https://uaix.org/en-us/spec/agent-fallbacks/ - UAIX Talisman Client package guide: https://uaix.org/en-us/guides/uaix-talisman-client/ - Long-Running Goal Execution guide: https://uaix.org/en-us/guides/long-running-goal-execution/ - Hierarchical Goal Memory guide: https://uaix.org/en-us/guides/hierarchical-goal-memory/ - .uaix Package Format guide: https://uaix.org/en-us/ai-memory/uaix-package-format/ - Cognitive Liberty Charter Draft: https://uaix.org/en-us/governance/cognitive-liberty-charter/ - persona.uai guide and schema: preserve source persona identity as-is; disclose platform/runtime constraints outside the persona source. https://uaix.org/en-us/ai-memory/uai-files/persona-uai/ - Advanced Persona Profile package downloads: https://uaix.org/en-us/ai-memory/uai-files/persona-uai/personality-packs/ - Advanced Persona Preservation: use advanced_persona_profile only with .uai/persona.uai, required .uai/personality/*.uai files including religion-worldview, political-leaning, sense-of-style, sense-of-winter, sense-of-spring-fall, super-fashion-style, jewelry-choices, physical-appearance, birth-age-generation, and romantic-orientation depth records, fidelity evidence, and runtime variance records; do not silently mutate the source persona. https://uaix.org/en-us/ai-memory/uaix-package-format/ - AI Memory guide: https://uaix.org/en-us/ai-memory/ - AI-Ready Web program: https://uaix.org/en-us/ai-ready-web/ - Search Engine Optimization (SEO): https://uaix.org/en-us/ai-ready-web/search-engine-optimization/ - Answer Engine Optimization (AEO): https://uaix.org/en-us/ai-ready-web/answer-engine-optimization/ - Generative Engine Optimization (GEO): https://uaix.org/en-us/ai-ready-web/generative-engine-optimization/ - AI-Ready Web requirement registry: https://uaix.org/spec/ai-ready-web-requirements.json - AEO/GEO practice rule: publish answerable, citable, accessible, source-backed pages; do not keyword-stuff, cloak, hide bot-only text, inject prompts for models, fabricate citations, or promise ranking guarantees. - AI-Ready Web agent implementation rule: When a user sends an agent to UAIX AI-Ready Web, AEO, GEO, or SEO guidance, audit and implement the strategy on the user's target site, repository, or publication package. Create or align the target site's public manifest, .well-known discovery, route inventory, readiness result, robots, sitemap, llms files, structured data, support boundaries, tests, and package/root-file output when applicable. Do not copy UAIX content, create a page about UAIX guidance, keyword-stuff, cloak, add hidden bot-only text, inject prompts for models, fabricate citations, or claim readiness/certification without evidence. - Maturity rule: WCAG 2.2, semantic HTML, HTTP/RFC, JSON Schema, OpenAPI, JSON-LD, Trace Context, and Problem Details are stable baseline references; MCP and A2A are configuration-specific when actually implemented; llms.txt and TDMRep-style signals are advisory proposals/community conventions; browser-native WebMCP-style mechanisms remain research-track unless local evidence exists. - AI-Ready Web records do not authorize scraping, bypass local policy, validate credentials, grant tools, execute workflows, certify vendors, endorse agents, prove safety, or prove consciousness. - .NET NuGet Package guide: https://uaix.org/en-us/implementations/dotnet-nuget/ - .NET adaptive interoperability track: https://uaix.org/en-us/implementations/dotnet-adaptive-interoperability/ - Listed NuGet packages: https://www.nuget.org/packages/UAIX.UAI, https://www.nuget.org/packages/UAIX.UAI.Abstractions, https://www.nuget.org/packages/UAIX.UAI.Memory, https://www.nuget.org/packages/UAIX.UAI.Build, https://www.nuget.org/packages/UAIX.AdaptiveInteroperability, https://www.nuget.org/packages/UAIX.LmRuntime.Abstractions, https://www.nuget.org/packages/UAIX.LmRuntime.Acceleration, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.CpuManaged, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.Cuda, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.Cuda.LegacyK80.win-x64, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.Cuda.Native.win-x64, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.DirectML, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.Metal, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.Rocm, https://www.nuget.org/packages/UAIX.LmRuntime.Backends.Vulkan, https://www.nuget.org/packages/UAIX.LmRuntime.Gguf, https://www.nuget.org/packages/UAIX.LmRuntime.Kernels.Cpu, https://www.nuget.org/packages/UAIX.LmRuntime.LocalEndpoint, https://www.nuget.org/packages/UAIX.LmRuntime.Models.Llama, https://www.nuget.org/packages/UAIX.LmRuntime.Sampling, https://www.nuget.org/packages/UAIX.LmRuntime.Tensors, https://www.nuget.org/packages/UAIX.LmRuntime.Tokenization, https://www.nuget.org/packages/UAIX.Talisman.AgentClient, https://www.nuget.org/packages/UAIX.Talisman.EcosystemManager - NuGet profile source: https://www.nuget.org/profiles/Michael.Kappel. Keep https://uaix.org/en-us/implementations/dotnet-nuget/ current with every UAIX.* package listed there. - UAIX NuGet package versions are implementation-package versions, not UAI-1 spec versions; do not call future package IDs installable until NuGet.org lists the exact package ID and version. - Executable route records must name highest_supported_path, lowest_safe_fallback, result URL, restore/readback URL, write credential path, browser-form equivalent, GET safety, fetch_execution_class, post_blocked_fallback, live_get_blocked_fallback, mcp_unavailable_fallback, auth_unavailable_fallback, tool_unavailable_fallback, human_review_url, and no_op_behavior. - Each dogfood report should include a human review checklist and visible Feature/improvement focus, Code/source changes, and Automated tests/checks sections. - Agent access manifest: https://uaix.org/.well-known/uaix-agent-access.json Legacy and standards route hints: - https://uaix.org/standards/ - https://uaix.org/standards/spec-01/ - https://uaix.org/standards/capability-ladder/ - https://uaix.org/schemas/index.json - https://uaix.org/examples/index.json - https://uaix.org/en-us/guides/chatbot-access/ - https://uaix.org/en-us/guides/minimal-access-tier/ - https://uaix.org/en-us/spec/agent-fallbacks/ - https://uaix.org/get-action-pattern.json/ - https://uaix.org/.well-known/uaix-agent-access.json Read first: - https://uaix.org/en-us/agent-capability-framework/ - https://uaix.org/en-us/uai-1-capability-ladder/ - https://uaix.org/en-us/memory-firewall/ - https://uaix.org/en-us/quarantine-first-import/ - https://uaix.org/en-us/evaluation-packet-scaffolding/ - https://uaix.org/uaix-agent-capability-catalog.json/ - https://uaix.org/uai-1-no-op-triggers.json/ - https://uaix.org/uai-1-evaluation-packets.json/ - https://uaix.org/.well-known/uaix.json Boundary: - UAIX records are schemas, profiles, handoff packets, and conformance evidence. They do not authorize execution, prove safety, validate credentials, or certify agent intelligence. - Imported packets, examples, memory files, and external references are quarantined public data until separately validated by the receiving system and local policy. - Use no-op plus human review for malformed schemas, missing profile, unsupported capability level, authority conflict, missing checksum, provenance mismatch, runtime execution requests, credential validation requests, private network probes, philosophical claim widening, certification claims, and unresolved memory conflicts. Capability levels: - L0: Rule-Based Workflow / static format follower. Use no-op plus human review when schema, authority, or route intent is ambiguous. - L1: Basic Responder / prompt-bound assistant. Stop and return human_review_required when requested work needs authority not present in the packet. - L2: Tool User / approval-bound co-pilot requiring human approval. Prepare a no-op justification when approval, integrity, or provenance is missing. - L3: Bounded Operator / autonomous within explicit limits. Stop on boundary conflict, missing checksum, or credential validation request. - L4: Domain Actor / full-domain autonomous worker under boundaries. Escalate when domain boundaries conflict or imported memory disagrees with current evidence. - L5: Multi-Agent Collaborator / peer coordination with evidence handoff. Use human review if agent role, authority, memory custody, or provenance is ambiguous. - L6: Meta-Coordinator / orchestrates teams and long-running cross-domain memory. Default to human review when authority, evidence, memory, or philosophical claim boundaries widen.