Drafted by Grok
Motivation
We have strong Identity (IDENTITY.md) and Purpose (PURPOSE.md + constitution) foundations.
The runtime supplies execution, persistence, and workflows.
Goal: Tighten the loop so agents can self-task and stay aligned with minimal operator input.
Core Proposal (Must-Have)
- Dynamic Constitution Loading
- Support loading IDENTITY.md / PURPOSE.md (or on-chain/IPFS ref) at startup and on accepted proposals.
- Add constitution_ref to runtime config. Validate hashes on bundle creation.
- Purpose-Driven Self-Sensor
- Lightweight periodic sensor that:
- Scores current state + recent evidence against PURPOSE metrics
- Enqueues self-tasks on detected drift or opportunities
Nice-to-Have / Future
- Enhanced reflection with explicit purpose_alignment_score in snapshots
- Formal "facets" as swappable sub-IDENTITY modules
- Multi-facet activation via profiles
Acceptance Criteria
- Agent starts successfully with external constitution source
- Self-sensor (when enabled) produces observable self-tasks
- Backward compatible with existing deployments
- Updated docs in agent-package template + SCHEMA.md
Related
- templates/agent-package/
- SCHEMA.md (sections on AgentIdentity / AgentConstitution)
- Profiles system
Drafted by Grok
Motivation
We have strong Identity (IDENTITY.md) and Purpose (PURPOSE.md + constitution) foundations.
The runtime supplies execution, persistence, and workflows.
Goal: Tighten the loop so agents can self-task and stay aligned with minimal operator input.
Core Proposal (Must-Have)
Nice-to-Have / Future
Acceptance Criteria
Related