公开记录项

公开记录项

此处说明当前公开记录、证据路径和支持边界。

  • 记录 UAIX-DOC-1490
  • 小路 /zh-cn/guides/uaix-uai-1-agent-skill/
  • 使用 规范公共记录

文件状态

公共标准页面 作为当前公共标准记录的一部分在 UAIX 上发布
代码
UAIX-DOC-1490
表面
公开记录项
使用权
公开且可链接

如何使用本页

此处说明当前公开记录、证据路径和支持边界。

UAIX.org publishes a UAIX / UAI-1 Agent Skill bundle for skills-compatible agents. The skill teaches agents how to create, validate, cite, and package UAI-1 evidence using UAIX’s public specification, validator, API routes, 接入套件, 一致性包, AI 记忆, and 项目交接 records. Agent Skills handle reusable procedure loading; UAI-1 handles portable public evidence.

What this page is for

Use this page when an agent or maintainer needs the reusable UAIX procedure bundle itself, rather than the broader agentic-harness strategy. The bundle is a reference SKILL.md folder with supporting files that a skills-compatible agent can load before producing UAIX evidence.

Download the reference UAIX / UAI-1 Agent Skill bundle. The current package is uaix-uai-1-agent-skill-v0.1.0.zip with SHA-256 2C654A1FBAA3646E4FB0B612EB036A2B3183EF7A9C3F4208FBCF12312F6DAB1D.

Layer split

Layer Job 边界
Agent Skill Load repeatable instructions, route maps, support boundaries, and packaging steps into a compatible agent session. Procedure guidance only; not a runtime, hosted importer, SDK, CLI, certification, endorsement, or official adapter.
UAI-1 Carry the reviewed public exchange packet, provenance, validator output, conformance material, and release or handoff evidence. 证据 for the named packet and scope; not a blanket conformance or certification claim.
AI 记忆 and 项目交接 Preserve accepted project facts, constraints, decisions, checks, receiver briefs, and next actions after review. Project memory; not UAI-1 conformance evidence by itself.

Use it when

证据 workflow

  1. 解析 current UAIX routes: use the public pages for human citations and the live API catalog, discovery, schemas, registry, examples, validator, 接入套件, 一致性包, and OpenAPI routes for automation.
  2. Choose the UAI-1 record shape: select the applicable profile and keep the schema, registry entry, example, and route URLs with the work product.
  3. Create the candidate packet: use readable keyed JSON for review, keep identity, target, delivery, trust, body, provenance, integrity, and extensions explicit when they apply, and redact sensitive material before validation.
  4. 验证 before claiming support: keep the exact checked packet, validator response, route URLs, timestamp, and named scope together.
  5. Cite public sources: cite the public page for prose claims, the machine route for automation, and the changelog or roadmap when the claim depends on release state or future support.
  6. Package the result: use the 接入套件 for a first-proof packet and the 一致性包 when release evidence needs to travel.
  7. Write back reviewed facts only: promote accepted facts into AI 记忆, 项目交接, release notes, tests, code, roadmap state, or machine artifacts after review.

Support boundary

  • The bundle is current as a reference procedure package with SKILL.md and supporting reference files.
  • It helps compatible agents resolve UAIX routes, select UAI-1 profiles, validate packets, preserve citations, and package first-proof or release evidence.
  • It is not an agent runtime, scheduler, orchestrator, SDK, CLI, certification authority, endorsement program, official adapter, hosted importer, automatic repository writer, automatic LLM Wiki sync service, package registry, authorization layer, or capability-package conformance profile.
  • Do not use it for AUiX, MedAI-UAIX, User-AI Experience research, or similarly named projects outside UAIX.org / 通用 Artificial Intelligence Exchange / UAI-1.
  • Do not place secrets, credentials, private customer data, hidden prompts, raw approvals, privileged operations, or unredacted runtime traces into UAI-1, AI 记忆, or 项目交接 records.