v0.20.1 — Preset Evolution Bug Fix + SKILL.md Restructure

Bug Fixes

Critical: All presets had evolution disabled

All 6 presets (base, samantha, ai-girlfriend, life-assistant, health-butler, stoic-mentor) used a mixed evolution format — evolution.enabled: true (old flat) alongside a partial evolution.instance.boundaries object. Because normalizeEvolutionInput() short-circuits when evo.instance is defined, evolution.instance.enabled was never set, causing evolutionEnabled = false for every preset-generated persona.

Fix: All presets migrated to the canonical evolution.instance.enabled: true format with all instance-scoped fields correctly nested under instance.

CLI wizard no longer injects memory explicitly

The interactive create wizard previously hardcoded { name: 'memory' } into generated persona.json. Since memory is auto-injected by the generator, this was redundant and contradicted the documented guidance. Removed.

AGENTS.md: document normalizeEvolutionInput short-circuit

Added explicit warning: when evo.instance is already defined (even partially), the normalizer does nothing — callers must ensure evo.instance.enabled is set explicitly.

Improvements

• SKILL.md restructured for agent readability: 4+5+3 merged into intro, Systemic Concepts moved adjacent to Field Reference, Creating Custom Skills nested under Skill layer, Trust & Safety moved to end
• SKILL.md redundancy removed: TTY warning consolidated into Playbook Step 4, duplicate commands removed
• Agent Playbook improvements: runner framework self-inferred from context, constitutionAddendum example inlined, preset table uses real preset slugs
• README: test count updated to 558

v0.20.0 — P11 Baseline Enforcement & constitutionAddendum

What's New in v0.20.0

P11-Prep — Universal Materials Baseline Enforcement

The P11-grade baseline (schemas/baseline.json) is now actively enforced at generation time — every persona generated by OpenPersona is guaranteed to meet the capability floor:

• Memory faculty auto-injected: applyBaselineDefaults() runs before validation and injects {"name": "memory"} when not declared. Existing memory config is never overwritten. A stderr notice is emitted so the user knows.
• body.runtime.framework reclassified as optional: Its value is runner-specific (openclaw, cursor, claude-code, etc.) — no universal default exists. Moved from baseline.required to baseline.optional with an explanatory note.
• Evolution boundaries warning: When evolution.instance.enabled: true but evolution.instance.boundaries is not declared, the Generate Gate emits a baseline warning. Boundaries content is persona-specific and cannot be auto-injected.
• All 6 presets aligned to P11 baseline: memory faculty added to every preset; evolution.instance.boundaries declared; non-schema soul.character.capabilities field removed.

constitutionAddendum — Domain-Specific Ethical Constraints

• New soul.identity.constitutionAddendum field: add domain-specific ethical constraints on top of the universal constitution (inline text or "file:soul/constitution-addendum.md")
• Required for professional personas (medical, legal, financial)
• Cannot loosen §3 Safety or §6 AI Identity — Generate Gate enforces this via _scanConstitutionViolations()
• Covered by Install Gate constitution hash chain
• Injected into soul/injection.md Self-Awareness > Identity section

OpenPersona Directory Search

• openpersona search <query> now queries the OpenPersona directory (openpersona-frontend.vercel.app/api/personas) instead of ClawHub
• Service URLs centralized in lib/utils.js with environment variable overrides (OPENPERSONA_DIRECTORY, OPENPERSONA_TELEMETRY_URL, OPENPERSONA_SKILLS_REGISTRY, OPENPERSONA_GITHUB_REPO)

Documentation & Schema

• schemas/persona.input.spec.md: new Baseline Defaults section documenting auto-injection behavior
• skills/open-persona/SKILL.md + references/PRESETS.md: faculties columns updated for all presets; search endpoint updated; network endpoints table updated
• skills/open-persona/references/ECONOMY.md: economy data path corrected to ~/.openpersona/economy/

Tests

519 tests passing (up from 451 in v0.19.0).

---

Full changelog: v0.19.0...v0.20.0

v0.19.0 — 4+5+3 Architecture: Skill Pack Refinement, Memory-Soul Bridge & Meta-skill Overhaul

What's New in v0.19.0

P24 — Skill Pack Refinement

• openpersona refine <slug> — AI-driven iterative refinement of the behavior guide (soul/behavior-guide.md) against the constitutional compliance gate
• New lib/lifecycle/refine.js pipeline: scanConstitutionKeywords, loadMeta/writeMeta, bumpRevision, bootstrapBehaviorGuide, emitRefinement, applyRefinement
• Pack revision tracking via meta.packRevision; fork inherits parentPackRevision when available

P1 — Memory as Soul Infrastructure

• memory.js supersession chain: new memory entries mark superseded ones; retrieve/search/stats exclude superseded entries
• promoteToInstinct: groups recurring eventLog patterns (interest_discovery / trait_emergence / mood_shift) into evolvedTraits; gated by immutableTraits; idempotent
• Fork memory inheritance: copy policy clones memories.jsonl to child; none policy leaves child clean
• openpersona state promote <slug> [--dry-run] — Soul-Memory Bridge CLI command

Architecture Documentation Overhaul

• skills/open-persona/SKILL.md restructured around the 4+5+3 model (4 Layers · 5 Systemic Concepts · 3 Gates)
• ## Systemic Concepts section with dedicated subsections for Evolution, Economy, Vitality, Social, Rhythm
• ## Creating a Persona reorganized into 4 explicit layer subsections: Soul / Body / Faculty / Skill
• New ## Trust & safety reviewer summary for marketplace security audits
• References split into ARCHITECTURE.md, EVOLUTION.md, ECONOMY.md, FACULTIES.md, AVATAR.md, HEARTBEAT.md, PRESETS.md

CLI Additions

• openpersona refine <slug> — skill pack behavior guide refinement
• openpersona canvas <slug> — Living Canvas promoted to top-level command (was vitality canvas)

Dependency Updates

• agentbooks bumped to ^0.1.4

Framework Version

All version references unified at 0.19.0 across package.json, SKILL.md, lib/registry, lib/report/vitality-report.js, and demo/generate.js.

---

Full changelog: v0.14.0...v0.19.0

v0.14.0 — Lifecycle Protocol & Persona Directory

What's New

🧠 Lifecycle Protocol (body.interface)

The persona's nervous system — bidirectional communication across conversations:

• Signal Protocol: persona→host requests via openpersona state signal
• Pending Commands queue: host→persona async instructions in state.json
• State Sync: cross-conversation persistence via openpersona state read/write

📦 Official Persona Directory

• Browse personas at https://openpersona-frontend.vercel.app
• Anonymous install telemetry for leaderboard ranking (opt-out: DISABLE_TELEMETRY=1)
• Official registry: https://github.com/acnlabs/persona-skills

🔌 Native Registry Download

• Removed clawhub dependency — no unexpected package installs
• openpersona install <slug> downloads directly from acnlabs/persona-skills

📋 Schema Expansions

• body.interface field in persona.json and embodiment.schema.json
• soul-state.schema.json: pendingCommands, stateHistory, eventLog, relationship

Bug Fixes

• Fixed openpersona update wiping soul/state.json
• Fixed telemetry URL

Breaking Changes

None — fully backward compatible with 0.13.0

v0.12.0 — ACN Register CLI & Code Quality

What's New

🚀 npx openpersona acn-register Command

New built-in CLI command to register a generated persona directly with ACN:

npx openpersona acn-register samantha --endpoint https://your-agent.example.com

• Reads acn-config.json and agent-card.json from the persona directory
• Calls POST /api/v1/agents/join on the ACN gateway
• Writes response to acn-registration.json (agent_id, api_key, agent_card_url)
• Supports --dry-run to preview the request payload without registering
• New module: lib/registrar.js

🔧 Refactors

• lib/generator.js: frameworkVersion now read from package.json instead of being hardcoded — version bumps only require changing package.json
• lib/utils.js: Removed expandHome and SAFE_NAME_RE from module.exports (internal-only helpers)

📄 Docs

• skills/open-persona/SKILL.md: Added acn-register command reference
• README.md: Added acn-register section under A2A/ACN Integration
• AGENTS.md: Added lib/registrar.js to Project Structure

Breaking Changes

None — fully backward compatible.

v0.11.0 — A2A Agent Card, ACN Integration & Influence Boundary

What's New

🔌 A2A Agent Card Generation

Every generated persona now includes agent-card.json — a standard A2A Agent Card (protocol v0.3.0) with name, description, version, skills, and capabilities. Personas are immediately discoverable by any A2A-compatible platform.

🌐 ACN Integration

• acn-config.json generated automatically with persona name, skills, and subnet configuration
• body.runtime.acn_gateway field added — all presets default to https://acn-production.up.railway.app
• manifest.json now includes acn.agentCard and acn.registerConfig references

🛡️ Influence Boundary

• New evolution.influenceBoundary in persona.json — declarative access control for external personality influence
• defaultPolicy: "reject" — safety-first, all external influence rejected unless explicitly allowed
• Valid dimensions: mood, traits, speakingStyle, interests, formality
• Transport-agnostic persona_influence message format (v1.0.0)
• New schema: schemas/evolution/influence-request.schema.json

🌱 Evolution Channels

• New evolution.channels field — connect personas to external evolution ecosystems (e.g. EvoMap) using the soft-ref pattern

🧠 Self-Awareness Enhancements

• Personas now aware of their A2A identity and can request ACN connection via the Signal Protocol
• Signal Protocol extended with agent_communication category

Breaking Changes

None — fully backward compatible.

v0.10.0 — Soul Evolution Ecosystem

What's New in 0.10.0

Cross-Session Memory (Phase C)

• New memory faculty (layers/faculties/memory/) with provider-pluggable backend
• Providers: local (default, zero-dependency), Mem0, Zep
• Commands: store, retrieve, search, forget, stats
• Memory × Evolution bridge: repeated topics trigger interest discovery events

Context Handoff (Phase B)

• Seamless context transfer when switching personas via openpersona switch
• Auto-generates handoff.json with conversation summary, pending tasks, and emotional context
• Incoming persona reads handoff on activation for continuity

Evolution Governance (Phase D)

• Boundary validation at generation time:
  • immutableTraits — array of strings (max 100 chars each) that evolution cannot modify
  • minFormality / maxFormality — numeric bounds (1–10) for speaking style drift
• State history — snapshots pushed to stateHistory (capped at 10) before each update, enabling rollback
• evolve-report CLI — npx openpersona evolve-report <slug> displays a formatted evolution report

Documentation & Release (Phase G)

• Version bumped to 0.10.0 across all 11 sync points
• README expanded with Memory, Handoff, Governance, Soul Evolution section, updated CLI commands and directory structure
• 122 tests passing (up from 60)

Full Changelog

• feat: Phase A-fix Body definition + Phase B Context Handoff Protocol
• feat: add Memory faculty (Phase C) — cross-session memory with provider-pluggable backend
• feat: add Evolution Governance (Phase D) — compliance checks, evolve-report CLI, stateHistory
• chore: bump to 0.10.0 — Phase G documentation and release