Agents and outward identity source context
This reference preserves cross-cutting context for agent definitions, runtime roles, definition synchronization, outward identity, and owner-representative conversation. It does not create requirements independently; binding statements are repeated in the applicable CRDs.
Source roles and limits
| Source | Role | Can establish | Cannot establish alone |
|---|---|---|---|
| Current PIOS framework | architecture authority | agent-definition/runtime/memory/activity separation, agent profiles as System information, canonical definition governance, runtime synchronization, PIOS-native versus independent roles, and owner authority | one agent hierarchy, model provider, runtime, file format, or public representative product |
| Historical PIOS Global at the selected revision | historical design evidence | personal-AI identity, runtime-role distinctions, external-peer context, outward identity modes, AI-Twin knowledge and audience boundaries, and human-authority intent | current architecture authority, universal permission policy, or current implementation |
Restricted application evidence RAS-01 | requirements and bounded static realization evidence | selected personal-AI identity/preferences and Role outward-intent controls already assessed | a canonical agent-definition model, runtime synchronization, AI-Twin realization, full outward-identity enforcement, tests, deployment, or production behavior |
Concept map
| Concept | Reusable meaning | Not equivalent to |
|---|---|---|
| Agent definition | Canonical portable description of a persistent agent | runtime process, runtime-local prompt, memory, activity log |
| Agent runtime | Process, machine, provider, or service executing an agent | canonical definition or owner-governed memory |
| Runtime role | Bounded functional position occupied during operation | persistent identity or permission grant |
| Personal-AI identity | Owner-governed presentation identity for the owner's private personal AI | runtime role or outward action mode |
| Outward identity | Identity/disclosure signal used for one external action or action class | permission to execute the action |
| Owner-representative conversation | External interaction through a bounded representative identity and knowledge scope | unrestricted access to Core or an entire AI-Twin product definition |
Cross-cutting boundaries
- Agent definition, runtime, owner-governed memory, and activity are separate. Replacing a runtime does not replace the definition or memory.
- Runtime roles describe functional responsibility and hierarchy at execution time; they are not agent identities and cannot create permissions by implication.
- Canonical agent definitions belong to governed System information. Runtime instruction files and provider-local state are deployed copies or operational artifacts whose synchronization status remains visible.
- Personal-AI identity, outward identity mode, and representative conversational behavior are separate outcomes.
- Acting under the owner's identity, as a disclosed named AI, or as a purpose-built representative produces materially different trust and disclosure conditions. The owner selects the applicable mode.
- A purpose-built outward AI representative uses a dedicated owner-curated knowledge scope and governed audience. Private Core information and private personal-AI conversations are not implicitly included.
- The named AI Twin concept composes agent definition, personal/outward identity, knowledge allocation, access, publication or sharing, retrieval, conversation, and governance capabilities. It is not treated as one indivisible Capability MLE.
Realization evidence
The selected restricted application has partial personal-AI identity and interaction-preference surfaces and a Role outward-intent selector assessed in earlier tranches. Repository-wide searches found no named agent-profile, runtime-target, or AI-Twin model at the selected revision. No canonical agent-definition synchronization or representative-conversation realization is therefore claimed.
Sequencing context
This tranche covers durable agent definition, deployment synchronization, runtime-role assignment, outward identity selection, and representative conversation. Agent work delegation, multi-agent orchestration, communication threads, model/provider gateways, and public-presence delivery remain separate capabilities or later candidates.