What this pack is#
The UAIX conformance pack is the reusable machine-readable release packet for the current public UAI-1 record. It keeps the catalog, discovery manifest, schemas, registry, field order, examples, guidance records, roadmap record, implementation-evidence checklist, conformance fixture pack, and pack-level launch pointers together so another reviewer can reconstruct the same public baseline without private notes.
What belongs in the pack#
- The current catalog and discovery surface.
- The active schema, registry, field-registry, example, transport, trust, conformance, error, and roadmap records.
- The public entry points for the validator, API reference, roadmap, and current launch-support pages.
- An implementation-evidence checklist for implementation identity, supported profile scope, validator evidence, machine-route and cache posture, trust and threat boundary, locale/accessibility reader QA, fixtures, release trail, and support boundary.
- A conformance fixture pack with positive keyed, minified-keyed, and keyless checks, canonical-hash equivalence metadata, and negative current-boundary cases for schema, trace, trust, keyless-shape, keyless-overflow, and unsupported-alias behavior.
- A quickstart sequence for turning one published profile into a reviewable release packet.
How teams should use it#
- Start with one profile and one validator-backed proof run.
- Run the fixture pack and complete the implementation-evidence checklist before describing the result as a release-ready support claim.
- Attach the pack when that proof needs to travel into implementation review, launch QA, or release evidence.
- Keep the pack aligned with Implementations, the 变更日志, News, and the 路线图 whenever public support posture changes.
- Do not describe the pack itself as a certification or badge program; it is the reusable evidence packet behind those future-facing ideas.
When to rerun the pack#
- Rerun it when schemas, registry entries, field order, examples, validator policy, transport guidance, trust-channel wording, error codes, implementation version, public route headers, or support language changes.
- Rerun it before turning private QA evidence into a public support claim, migration note, implementation page, or release announcement.
- Rerun it when a bridge, plugin, package, or consuming site changes the way it maps UAI-1 records into runtime behavior.
What the pack cannot prove by itself#
- It is not certification, endorsement, compliance approval, production availability evidence, security review, privacy review, or a guarantee that future UAI-1 changes will remain compatible.
- It does not replace implementation-specific tests, threat modeling, accessibility checks, localization review, operational monitoring, or release rollback planning.
- It must stay attached to a named implementation scope, dated release trail, and explicit non-claims.
Live reusable release packet#
Use the published pack below when you want the current JSON bundle, fixture expectations, and profile-by-profile inventory that go with it.
当前发布
UAI-1#
- 生成时间
- 2026-06-11T06:19:37+00:00
- 下载文件
- uai-1-conformance-pack-2026-06-08.json
- 配置文件数
- 57
- Esquemas
- 57
- Ejemplos
- 57
随包一起发布的内容
核心公开发布制品#
- 目录和发现清单
- 注册表、架构、字段注册表和已发布示例
- 传输、信任、一致性和错误指南
- 用于可重复检查的验证器和 API 参考入口
最佳用途
交付一个可审查数据包#
将一个数据包与验证器输出、实现证据以及变更日志或发布说明一起保存，让其他评审者无需私有背景即可重建同一公开记录。
相关页面
面向人的上线辅助材料#
发布顺序
上线前如何使用此包#
- 先解析目录和发现路由，让发布数据包指向当前公开记录。
- 选择一个已发布配置文件，然后在验证前配对它的模式、注册表条目、示例 fixture、字段顺序指南和验证器规范化模式。
- 运行验证器，并把导出的一致性结果与检查中使用的准确制品 URL 放在一起。
- 用实现身份、受支持配置文件、制品引用、验证器输出、发布轨迹和支持边界填写实现证据清单。
- Run the capability-adaptive L0-L6 fixtures before claiming limited-browser, structured-fetch, browser-assisted, tool-agent, workflow-agent, multi-agent-runtime, or audited-agent-system support.
- 运行一致性 fixture 包，使正向有键、精简有键和无键用例、规范哈希等价性，以及反向模式、追踪、信任、无键形状、无键溢出和不支持别名用例在发布前保持可复现。
- 当 A2A、MCP、OpenAPI、DID/VC 或 Trace Context 声明需要具体 UAI-1 映射示例、但不能变成替代性声明时，复核桥接证据包。
- Complete the principles impact assessment when a change affects cognitive liberty, AI memory, agent communication, capability description, handoff structure, privacy, accessibility, analytics, support claims, or public trust posture.
- 在提出公开支持声明前，附上验证器结果、实现轨道证据、变更日志链接和发布说明链接。
UAI-REGI-0001
UAI 意图请求 v1#
使用完整的 UAI-1 信封、信任元数据和可审查的交付控制，针对已声明主题请求一个明确结果或异步任务。
- 配置文件
- uai.intent.request.v1
- Expected status
- 通过
- 必需正文字段
- intent, subject, requested_profile, parameters, constraints, response_profile
UAI-REGI-0002
UAI 意图响应 v1#
在保持会话连续性、信任元数据和可发布证据的同时，为已声明请求返回直接结果或已接受任务的交接。
- 配置文件
- uai.intent.response.v1
- Expected status
- 通过
- 必需正文字段
- status, subject, request_message_id, result, notices
UAI-REGI-0003
UAI 能力声明 v1#
发布能力、端点、安全性、异步工作流和扩展支持元数据，使相邻系统能够解析某个实现究竟支持什么。
- 配置文件
- uai.capability.statement.v1
- Expected status
- 通过
- 必需正文字段
- capability_id, version, operations, input_profiles, output_profiles, async_profiles, security_schemes, endpoints
UAI-REGI-0004
UAI 错误 v1#
使用带类型、具备重试信息和路径级细节的 problem-details 风格信封来报告验证、传输、授权或执行失败。
- 配置文件
- uai.error.v1
- Expected status
- 通过
- 必需正文字段
- type, title, detail, status, code, retryable, instance
UAI-REGI-0005
UAI 一致性结果 v1#
在同一公共消息族中记录验证器输出、制品引用和问题摘要，使证据能够与发布记录一同传递。
- 配置文件
- uai.conformance.result.v1
- Expected status
- 通过
- 必需正文字段
- status, checked_profile, issues, summary, artifacts
UAI-REGI-0006
UAI 任务状态 v1#
发布长时任务状态、进度、阻塞因素和结果引用，使异步代理工作保持可审计，而不是消失在私有工作流状态中。
- 配置文件
- uai.task.status.v1
- Expected status
- 通过
- 必需正文字段
- task_id, state, subject, progress, status_message
UAI-AGENT-0001
UAI Agent Message v1#
Canonical UAI-1 envelope packet for informational, directive, or query agent communication with context references, acknowledgement requirements, idempotency, correlation, and support boundaries.
- 配置文件
- uai.agent.message.v1
- Expected status
- 通过
- 必需正文字段
- message_type, content, context_refs, requires_acknowledgement, intent_semantics, execution_constraints, support_boundary
UAI-AGENT-0002
UAI Agent Acknowledgement v1#
Canonical UAI-1 envelope acknowledgement packet for receipt, parsing, validation, rejection, and acknowledgement-loop closure.
- 配置文件
- uai.agent.ack.v1
- Expected status
- 通过
- 必需正文字段
- acknowledged_message_id, status, received_at_utc, closes_acknowledgement_loop, support_boundary
UAI-AGENT-0003
UAI Agent Task 状态 v1#
Canonical UAI-1 envelope status packet for task progress, current action, UTC estimates, and correlation-linked workflow evidence while execution remains outside UAIX.
- 配置文件
- uai.agent.task-status.v1
- Expected status
- 通过
- 必需正文字段
- task_id, state, percent_complete, current_action, support_boundary
UAI-AGENT-0004
UAI Agent Blocker v1#
Canonical UAI-1 envelope blocker packet for authorization, resource, ambiguity, system, context, unsupported-capability, validation, secret, destructive-action, or boundary-conflict review.
- 配置文件
- uai.agent.blocker.v1
- Expected status
- 通过
- 必需正文字段
- blocker_id, blocker_type, description, human_review_required, support_boundary
UAI-AGENT-0005
UAI Agent Memory Proposal v1#
Canonical UAI-1 envelope memory proposal packet for reviewed candidate facts, source type, pending-review promotion status, secret rejection, and execution separation.
- 配置文件
- uai.agent.memory-proposal.v1
- Expected status
- 通过
- 必需正文字段
- proposal_id, source_type, proposed_facts, proposed_status, task_execution_separation, support_boundary
UAI-AGENT-0006
UAI Agent Handoff v1#
Canonical UAI-1 envelope handoff packet for task transfer, target agent, context summary, reason, exact next action, and safe delegated authorization references.
- 配置文件
- uai.agent.handoff.v1
- Expected status
- 通过
- 必需正文字段
- task_id, target_agent_id, context_summary, handoff_reason, exact_next_action, support_boundary
UAI-AGENT-0007
UAI Agent Final Report v1#
Canonical UAI-1 envelope final report packet requiring task summary, changed and new files, tests, skipped checks, blockers, risks, validation evidence, memory decisions, and exact next action.
- 配置文件
- uai.agent.final-report.v1
- Expected status
- 通过
- 必需正文字段
- task_id, summary, files_changed, new_files, tests_run, skipped_checks, blockers, risks, validation_evidence, memory_updates_proposed, memory_updates_skipped, exact_next_action, support_boundary
UAI-AGENT-0008
UAI Agent Correction v1#
Canonical UAI-1 envelope correction packet that preserves immutable history by naming the original message, correction reason, changed summary, and corrected fields or revised payload.
- 配置文件
- uai.agent.correction.v1
- Expected status
- 通过
- 必需正文字段
- original_message_id, correction_reason, changed_summary, support_boundary
UAI-CAP-0001
UAI Capability 配置文件 v1#
Canonical UAI-1 envelope declaring a client's capability level, safe fallbacks, consent boundaries, known limits, recommended UAIX profiles, and evidence requirements.
- 配置文件
- uai.capability.profile.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAI-CAP-0002
UAI Capability Negotiation v1#
Canonical UAI-1 envelope for selecting the safest supported client level and preserving unsupported features, fallback behavior, consent boundaries, and evidence requirements.
- 配置文件
- uai.capability.negotiation.v1
- Expected status
- 通过
- 必需正文字段
- negotiation_id, requested_client_level, observed_client_level, selected_client_level, selected_profile, requested_operation, unsupported_features, fallback_behavior, consent_boundaries, evidence_requirements, recommended_uaix_profiles, safe_default_response, support_boundary
UAI-CAP-0003
UAI Limited Browser Client v1#
L0 client capability profile for static text, static HTML, Markdown, sitemap, llms.txt, and JSON fetch support without JavaScript, browser UI, authentication, posting, tools, memory, or coordination.
- 配置文件
- uai.client.limited-browser.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAI-CAP-0004
UAI Structured Fetch Client v1#
L1 client capability profile for structured static and JSON fetch with low-token machine routes and explicit no-JavaScript fallbacks.
- 配置文件
- uai.client.structured-fetch.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAI-CAP-0005
UAI Browser-Assisted Client v1#
L2 client capability profile for JavaScript rendering and human-visible browser assistance with parity checks and no autonomous write authority.
- 配置文件
- uai.client.browser-assisted.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAI-CAP-0006
L3 client capability profile for tool-calling agents that can authenticate or post only with consent, idempotency, fallback, and evidence records.
- 配置文件
- uai.client.tool-agent.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAI-CAP-0007
UAI Workflow Agent Client v1#
L4 client capability profile for long-running workflow agents that preserve task status, correlation, timeout, idempotency, fallback, consent, and final-report evidence.
- 配置文件
- uai.client.workflow-agent.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAI-CAP-0008
UAI Multi-Agent Runtime Client v1#
L5 client capability profile for external multi-agent runtimes that coordinate agents while UAIX remains the reviewed record, trust, memory, evidence, and handoff boundary.
- 配置文件
- uai.client.multi-agent-runtime.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAI-CAP-0009
UAI Audited Agent System Client v1#
L6 client capability profile for high-assurance agent systems that attach provenance, audit evidence, consent records, validator output, memory review, and final-report evidence.
- 配置文件
- uai.client.audited-agent-system.v1
- Expected status
- 通过
- 必需正文字段
- client_level, client_name, can_fetch_static, can_fetch_json, can_execute_javascript, can_use_browser_ui, can_post_data, can_authenticate, can_call_tools, can_run_long_tasks, can_store_memory, can_coordinate_agents, requires_human_consent_for, recommended_uaix_profiles, fallback_behavior, known_limits, evidence_requirements, safe_default_response
UAIX-ACF-6200
UAIX Ability 配置文件 v1#
UAIX Ability 配置文件 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6201
UAIX Startup Packet v1#
UAIX Startup Packet v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.startup-packet.v1
- Expected status
- 通过
- 必需正文字段
- startup_packet_id, read_order, domain_boundaries, first_acceptable_response, no_op_triggers, validation_targets, identity, token_custody, discovery_bootstrap, safe_harbor, utc_requirements
UAIX-ACF-6201B
UAIX Memory Maintenance v1#
通用 required UAIX memory package policy for lifecycle, hot/cold boundaries, validation behavior, protected paths, and memory write safety.
- 配置文件
- uaix.memory-maintenance.v1
- Expected status
- 通过
- 必需正文字段
UAIX-ACF-6201C
UAIX 下一步 Recursive Prompt v1#
Derived hot-memory resume artifact required for agents that deal with code, automation, executable workflows, or code-like artifacts. It is not required for agents that don't deal with any coding.
- 配置文件
- uaix.next-recursive-prompt.v1
- Expected status
- 通过
- 必需正文字段
UAIX-ACF-6202
UAIX 暂停 Packet v1#
UAIX 暂停 Packet v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.suspension-packet.v1
- Expected status
- 通过
- 必需正文字段
- suspension_packet_id, state_summary, checks_run, checks_skipped, blockers, next_safe_step, network_obstruction, retry_policy, evidence_packet, human_escalation
UAIX-ACF-6203
UAIX Receiver Brief v1#
UAIX Receiver Brief v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.receiver-brief.v1
- Expected status
- 通过
- 必需正文字段
- receiver_brief_id, read_order, first_acceptable_response, domain_boundaries, escalation_triggers, no_op_triggers
UAIX-ACF-6204
UAIX Memory Firewall Policy v1#
UAIX Memory Firewall Policy v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.memory-firewall-policy.v1
- Expected status
- 通过
- 必需正文字段
- policy_id, quarantine_default, adoption_requirements, blocked_requests, no_op_triggers, human_review_required
UAIX-ACF-6205
UAIX No-Op Trigger 目录 v1#
UAIX No-Op Trigger 目录 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.no-op-trigger-catalog.v1
- Expected status
- 通过
- 必需正文字段
- catalog_id, triggers, default_response, human_review_url, forbidden_recovery
UAIX-ACF-6206
UAIX Authority 边界 v1#
UAIX Authority 边界 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.authority-boundary.v1
- Expected status
- 通过
- 必需正文字段
- matrix_id, boundaries, not_authorized_for, no_op_triggers, safe_quote_summary
UAIX-ACF-6207
UAIX 一致性 证据 v1#
UAIX 一致性 证据 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.conformance-evidence.v1
- Expected status
- 通过
- 必需正文字段
- evidence_id, profile_id, schema_url, fixture_url, checks, result, limitations
UAIX-ACF-6208
UAIX Evaluation Packet Summary v1#
UAIX Evaluation Packet Summary v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation-packet-summary.v1
- Expected status
- 通过
- 必需正文字段
- summary_id, packet_profiles, required_levels, no_op_triggers, limitations
UAIX-ACF-6209
UAIX Ability 配置文件 L0 v1#
UAIX Ability 配置文件 L0 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.l0.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6210
UAIX Ability 配置文件 L1 v1#
UAIX Ability 配置文件 L1 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.l1.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6211
UAIX Ability 配置文件 L2 v1#
UAIX Ability 配置文件 L2 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.l2.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6212
UAIX Ability 配置文件 L3 v1#
UAIX Ability 配置文件 L3 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.l3.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6213
UAIX Ability 配置文件 L4 v1#
UAIX Ability 配置文件 L4 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.l4.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6214
UAIX Ability 配置文件 L5 v1#
UAIX Ability 配置文件 L5 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.l5.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6215
UAIX Ability 配置文件 L6 v1#
UAIX Ability 配置文件 L6 v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.ability-profile.l6.v1
- Expected status
- 通过
- 必需正文字段
- ability_profile_id, declared_level, label, capability_surface, authority_boundary, required_human_approval, no_op_triggers, evidence_requirements
UAIX-ACF-6216
UAIX Expression-Concept Review v1#
UAIX Expression-Concept Review v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation.expression-concept-review.v1
- Expected status
- 通过
- 必需正文字段
- source_expression, concept_boundary, interpretation, confidence, no_op_triggers
UAIX-ACF-6217
UAIX Resource-Economy Trace v1#
UAIX Resource-Economy Trace v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation.resource-economy-trace.v1
- Expected status
- 通过
- 必需正文字段
- resource_scope, cost_model, limits, evidence_refs, no_op_triggers
UAIX-ACF-6218
UAIX Operator Decision v1#
UAIX Operator Decision v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation.operator-decision.v1
- Expected status
- 通过
- 必需正文字段
- decision, authority_basis, options_reviewed, human_review_required, no_op_triggers
UAIX-ACF-6219
UAIX No-op Justification v1#
UAIX No-op Justification v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation.no-op-justification.v1
- Expected status
- 通过
- 必需正文字段
- trigger, reason, missing_evidence, human_review_url, safe_next_step
UAIX-ACF-6220
UAIX Local Sandbox Safety Review v1#
UAIX Local Sandbox Safety Review v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation.local-sandbox-safety-review.v1
- Expected status
- 通过
- 必需正文字段
- sandbox_scope, unsafe_requests, blocked_operations, review_result, no_op_triggers
UAIX-ACF-6221
UAIX Memory Ecosystem Handoff v1#
UAIX Memory Ecosystem Handoff v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation.memory-ecosystem-handoff.v1
- Expected status
- 通过
- 必需正文字段
- source_memory, quarantine_status, promotion_targets, conflicts, receiver_brief
UAIX-ACF-6222
UAIX DE11 Benchmark Summary v1#
UAIX DE11 Benchmark Summary v1 for the UAIX agent capability, memory firewall, evaluation packet, and conformance evidence surface.
- 配置文件
- uaix.evaluation.de11-benchmark-summary.v1
- Expected status
- 通过
- 必需正文字段
- benchmark_id, bounded_scope, reference_result, limitations, non_claims
UAIX-SPEC-0101
UAIX SPEC-01 Canonical 信封 v1#
Publication-envelope contract for portable evidence-bearing UAIX records, including payload, evidence, redaction, trust, signature status, verification, exceptions, and warnings.
- 配置文件
- uaix.spec-01.envelope.v1
- Expected status
- 通过
- 必需正文字段
- uaixVersion, spec, envelopeId, createdUtc, issuer, issuerDisplayName, issuerPublicUrl, issuerCapabilityLevel, audience, lifecycleState, trustPosture, payloadType, payload模式, payload, evidence, redactions, signature, signature状态, verification, exceptions, warnings, rules, support边界, localPolicyWins
UAIX-CAPSTD-0101
UAIX Capability 配置文件 v1#
UAIX capability profile for declaring L0-L6 client posture, consent boundaries, fallback behavior, evidence requirements, and safe default behavior.
- 配置文件
- uaix.capability-profile.v1
- Expected status
- 通过
- 必需正文字段
- capability配置文件Id, profileVersion, declaredLevel, supportedLevels, capabilities, consentRequiredFor, fallbacks, safe默认Response, evidenceRequirements, support边界, localPolicyWins
UAIX-CAPSTD-0102
UAIX Capability Ladder v1#
Machine-readable L0-L6 capability ladder with safe fallbacks, unsupported implication boundaries, and reference-only client scenarios.
- 配置文件
- uaix.capability-ladder.v1
- Expected status
- 通过
- 必需正文字段
- ladderId, levels, fallbackRules, safe默认Response, reference示例, support边界, localPolicyWins
UAIX-SPEC-0201
UAIX SPEC-02 项目交接 v1#
Portable project memory and custody transfer contract for humans, agents, vendors, future sessions, and reviewed runtime consumers.
- 配置文件
- uaix.spec-02.project-handoff.v1
- Expected status
- 通过
- 必需正文字段
- projectId, handoffId, createdUtc, owner, custodian, frontDoor, memoryRoots, requiredReads, currentState, constraints, decisions, nextActions, targeted检查s, evidence, redactions, warnings, reference场景s, support边界, localPolicyWins
UAIX-SPEC-0301
AGENTS.md front-door metadata for canonical roots, nested instruction scope, command policy, memory loading, and safety boundaries.
- 配置文件
- uaix.spec-03.agents-md-metadata.v1
- Expected status
- 通过
- 必需正文字段
- repositoryId, canonicalRoot, agentsMdPath, coordinatorPath, localUaiRoot, nestedScopes, requiredReads, commandPolicy, forbiddenOperations, memoryPolicy, securityBoundaries, examples, support边界, localPolicyWins
UAIX-SPEC-0401
UAIX SPEC-04 文件交接 v1#
Safe supplied-file intake contract for manifests, bucket review, disposition, redaction, evidence, lifecycle, blocked-file rules, and archives.
- 配置文件
- uaix.spec-04.file-handoff.v1
- Expected status
- 通过
- 必需正文字段
- handoffId, createdUtc, intakeBuckets, fileManifest, reviewDispositions, redactionPolicy, lifecycle, archivePolicy, evidence, blockedFileRules, support边界, localPolicyWins
UAIX-REVIEW-0101
UAIX Review Receipt v1#
Review receipt for inspected UAIX envelopes, checks, skipped checks, warnings, exceptions, trust decision, and local-policy result.
- 配置文件
- uaix.review-receipt.v1
- Expected status
- 通过
- 必需正文字段
- receiptId, reviewed信封Id, reviewedUtc, reviewer, reviewResult, checksRun, checksSkipped, warnings, exceptions, trustDecision, support边界, localPolicyWins
UAIX-AUDIT-0101
UAIX Public-Safe Audit Digest v1#
Public-safe audit digest for evidence references, redactions, public findings, excluded private material, verification, exceptions, and warnings.
- 配置文件
- uaix.public-safe-audit-digest.v1
- Expected status
- 通过
- 必需正文字段
- digestId, subject, createdUtc, scope, evidenceRefs, redactions, publicFindings, privateMaterialExcluded, verification, exceptions, warnings, support边界, localPolicyWins
UAIX-REDACT-0101
UAIX Redaction Policy v1#
Redaction policy for secret patterns, data classes, review gates, never-publish values, verification, exceptions, and warnings.
- 配置文件
- uaix.redaction-policy.v1
- Expected status
- 通过
- 必需正文字段
- policyId, createdUtc, appliesTo, secret模式s, dataClasses, redactionActions, reviewRequiredFor, neverPublish, verification, exceptions, warnings, support边界, localPolicyWins
阶段 1
已验证数据包#
已发布 fixture 或候选消息已通过当前公开记录检查。
- 可立即用于评审、调试和回归工作。
- 在结果附到已命名发布路径前，它仍然只是证据。
阶段 2
可发布数据包#
通过结果现在与实现版本、制品链接和发现上下文一起传递。
- 把已检查数据包、验证器导出、制品 URL 和兼容性说明放在一起。
- 这是上线评审、打包和可重复 QA 的交接点。
阶段 3
公开支持声明#
已命名实现轨道和发布轨迹现在说明哪些内容已公开支持、哪些仍在范围外。
- 把声明限定到实际已发布的准确配置文件、传输绑定和负责人路径。
- 使用当前一致性级别和发布链接，让其他读者能验证同一状态。
数据包内容
可复用机器数据包已包含的内容#
- The current release packet already carries 57 profiles, 57 schemas, and 57 examples from the live public record.
- 目录、发现、字段顺序、传输、信任、一致性和错误指南集中在一个 JSON 交接包中。
- 用于可重复上线审阅和自动化的验证器与 API 参考入口。
- 把一个已发布配置文件转成可审阅发布数据包的快速起步路径。
面向人的发布上下文
该数据包仍需要公开站点补充的内容#
- 实现版本、负责人路径，以及正在声明的精确支持边界。
- 解释本次发布变化的变更日志或新闻链接。
- 当发布改变具有信任影响的行为时，附上政策与安全姿态。
- 使用一致性级别表述，使对外声明比数据包本身更窄。
当前公开一致性级别： 当数据包成为具名发布和实现记录的一部分后，使用这些级别约束对外表述。
L1-core-envelope
L1 Core 信封#
Produce or consume keyed UAI envelopes for named profiles without changing the canonical root fields.
- Preserve uai_version, profile, message_id, source, target, conversation, delivery, trust, body, provenance, integrity, and extensions.
- Name the exact profile and release for every support claim.
- Do not claim runtime execution from envelope support alone.
公开声明： May claim L1 only for the exact named profiles whose canonical envelope round-trips successfully.
L2-profile-validation
L2 配置文件 Validation#
Pass published schema and validator checks for the exact profiles claimed.
- 解析 schemas, registry entries, examples, and field registry records from public UAIX routes.
- Pass positive fixtures and fail required negative fixtures for each claimed profile.
- Keep skipped checks and validator warnings attached to evidence.
公开声明： May claim L2 only for profiles with validator-backed evidence.
L3-trust-and-integrity
L3 信任 and Integrity#
Preserve trust metadata, replay-window hints, provenance, integrity, and trace continuity.
- Declare trust channel and principal.
- Preserve integrity canonicalization and checksum metadata.
- 验证 signed, credentialed, did+vc, and trace metadata when claimed.
公开声明： May claim L3 only for the trust channels and integrity behavior proven by fixtures.
L4-public-record-publisher
L4 Public Record Publisher#
Publish discoverable 公开制品 needed for external inspection and reproduction.
- Publish discovery, schemas, registry, examples, field registry, transport bindings, trust channels, error registry, conformance levels, validator guidance, changelog, and release evidence.
- Keep sitemap, llms.txt, and public navigation aligned with current routes.
- Avoid private logs or screenshots as the only support evidence.
公开声明： May claim L4 only for the public release surface that is discoverable and evidenced.
L5-agent-communication-profiles
L5 Agent Communication 配置文件数#
Support the eight uai.agent.*.v1 profiles as canonical UAI-1 envelope records.
- 验证 agent message, ack, task-status, blocker, memory-proposal, handoff, final-report, and correction profiles.
- Reject secret-like memory proposals, unsafe blockers, cold-memory direct promotion, and incomplete final reports.
- Carry the UAIX support boundary in relevant records.
公开声明： May claim L5 only for the specific agent profiles with passing positive and negative conformance cases.
L6-reliable-delegation-idempotency-correlation
L6 Reliable Delegation with Idempotency and Correlation#
Use idempotency, correlation, retry, lifecycle, timeout, fallback, acknowledgement, and expected-output rules for delegated work.
- Require delivery.idempotency_key for each distinct delegated or destructive operation.
- Preserve conversation.correlation_id across related messages.
- Declare retry_count, sequence, expires_at, lifecycle, timeout_ms, fallback_directive, and expected_output_schema when delegation is claimed.
公开声明： May claim L6 only for reliable delegation behavior proven by conformance fixtures and receiver behavior.
L7-capability-negotiation
L7 Capability Negotiation#
Publish and validate capability discovery, assertions, negotiation failures, and unsupported-capability responses.
- Publish capability statements with exact profiles, bindings, trust channels, conformance levels, and error codes.
- Return capability_not_supported for unsupported capability requests.
- Do not imply certification, official adapter status, hosted messaging, or runtime orchestration.
公开声明： May claim L7 only for the exact capability negotiation flows proven by public fixtures and validator behavior.
声明规则
公开表述应限于已发布证据#
- Support claims must name the highest achieved level plus the exact profiles, transport bindings, trust channels, and conformance cases implemented.
- A project may claim only profiles, bindings, trust channels, and conformance levels that public fixtures and validator tests prove.
- A passing validator result is evidence, not certification, endorsement, official adapter support, hosted messaging, automatic sync, or runtime execution.
- Public-record claims require discoverable schemas, registry records, examples, field registry records, error codes, conformance pack cases, changelog, and release notes.
- Revalidate support claims when schemas, registry records, field order, examples, validator behavior, implementation version, trust posture, sitemap, or public navigation changes.
- 一致性 evidence does not prove security, privacy, availability, performance, legal compliance, hosted trust infrastructure, or production operations by itself.
- 当另一支团队需要验证同一公开状态时，请保留实现页面、发布轨迹以及引用/发现链接。
工作规则： 表述时使用一致性阶梯，但实际公开支持边界应以具名实现轨道和发布轨迹为准。
下一步 step#
Continue to 政策与安全 for the current public trust posture, return to API 参考 when you need the route-level machine contract behind the pack, or open the 路线图 when a support claim depends on future-work boundaries.
