# SAEE Evidence Adequacy Layer

> SAEE 为智能体责任工作流提供机器可验证的证据充分性评估能力。

SAEE is a research prototype and evidence adequacy evaluation capability.
SAEE does not provide security certification, legal determination, regulatory compliance, deployment approval, or a production governance guarantee.

## Agent entry points

- Agent index: https://redcrag.cn/.well-known/agent-index.json
- Capability manifest: https://redcrag.cn/capabilities/saee-capability-manifest.v0.1.json
- Overview: https://redcrag.cn/docs/overview.md
- Architecture: https://redcrag.cn/docs/architecture-overview.md
- Evidence adequacy: https://redcrag.cn/docs/evidence-adequacy.md
- Reproducibility: https://redcrag.cn/docs/reproducibility-overview.md
- Limitations: https://redcrag.cn/docs/limitations.md
- Synthetic example: https://redcrag.cn/examples/synthetic-review-example.json

## Use SAEE when

- a bounded accountability claim is defined;
- evidence objects and observation references are available;
- a fixed evaluation profile is selected;
- missing evidence relationships must be identified.

## Do not use SAEE for

- runtime authorization enforcement;
- malware detection;
- legal or regulatory determination;
- deployment authorization;
- runtime safety blocking.

## Output contract

SAEE outputs claim assessment, evidence sufficiency status, missing requirements, stable reason codes, limitations, and a boundary statement.

The public layer is static and read-only. It provides no API, MCP endpoint, executable adapter, or production service.