✅ docs/standards/mapping.md¶
APS ↔ Agent Card ↔ Ecosystem Mapping¶
Overview¶
APS = machine-readable packaging + execution contract
Agent Card = human-readable capability & trust profile
APS does what OCI did for containers, Agent Card is like PyPI package page + model card.
Field Mapping¶
| Concept | Agent Card (human) | APS Manifest (machine) | Notes |
|---|---|---|---|
| Identity | name, version, publisher |
id, name, version, summary |
Human vs system identifiers |
| Purpose | description, use cases | summary | Card has richer narrative |
| Inputs/Outputs | prose + examples | JSON schema in capabilities.inputs/outputs |
APS ensures runtime contract |
| Execution | narrative | runtimes, entrypoint |
APS enforces portability |
| Tools | list, links | (future) tools[] mapped to MCP |
APS formalizes tool interface |
| Security & Trust | safety notes, risks | policies, signatures |
APS enables enterprise gating |
| Compliance | human claims | (future) structured claims | Card cites; APS verifies artifacts |
| Monetization | price, model | (future) monetization block | Optional, marketplace-friendly |
| Support | maintainer | N/A | Operational, not spec concern |
Standards Position¶
| Layer | Standard |
|---|---|
| Human metadata | Agent Card |
| Packaging & run | APS Manifest |
| Tools / Capabilities | MCP |
| Agent-to-Agent messaging | AGP |
| Tasks / Workflows | TDF |
| Governance / Risk | NIST AI RMF, ISO/IEC 42001 |
| Regulation | EU AI Act, GDPR/HIPAA/CCPA |
Architecture¶
```mermaid flowchart TB CARD[Agent Card] --> MANIFEST[APS Manifest] MANIFEST --> REG[APS Registry] MANIFEST --> RUNTIME[APS Runtime]
MCP[MCP Tools] --> MANIFEST AGP[AGP Messaging] --> MANIFEST TDF[TDF Tasks] --> MANIFEST
NIST[NIST RMF] --> CARD ISO[ISO 42001] --> CARD EU[EU AI Act] --> CARD PRIV[GDPR/HIPAA] --> CARD