arkaos 2.56.0 → 2.57.0

package/VERSION

@@ -1 +1 @@
-2.56.0
+2.57.0

package/departments/kb/skills/persona-build/SKILL.md

@@ -1,7 +1,10 @@
 ---
 name: kb/persona-build
 description: >
-  Build a persona from content: extract DISC, Enneagram, Big Five, MBTI, mental models.
+  Build or refine a callable persona from KB content. Extracts beliefs, voice
+  signature, expertise domains, decision patterns, and the 4-framework
+  behavioral DNA (DISC + Enneagram + Big Five + MBTI). Produces a callable
+  advisor YAML.
 allowed-tools: [Read, Write, Edit, Bash, Grep, Glob, Agent, WebFetch, WebSearch]
 ---
 
@@ -21,12 +24,188 @@ does not replace the vault.
 
 # Persona Build — `/kb persona <name>`
 
-> **Agent:** Clara (Knowledge Director) | **Framework:** Persona DNA Framework (4 frameworks)
+> **Lead:** Clara (Knowledge Director) | **Cross-dept:** Pedro (Research Analyst) + Eduardo (Copy Director) | **Frameworks:** Persona DNA (DISC + Enneagram + Big Five + MBTI) + Voice Pattern Extraction + Source-Cited Belief Inventory
 
-## What It Does
+## What ships
 
-Build a persona from content: extract DISC, Enneagram, Big Five, MBTI, mental models.
+A production callable persona in 7 deliverables:
 
-## Output
+1. **Source inventory** — every input file/URL tagged by type + date
+2. **Belief inventory** — core beliefs with source citations per belief
+3. **Voice pattern** — N-gram signature + recurring phrases + structural rhythm + no-go words
+4. **Expertise map** — deep / surface / no-go domain classification
+5. **Decision pattern catalogue** — characteristic evaluation patterns with cited examples
+6. **4-framework Behavioral DNA** — DISC + Enneagram + Big Five + MBTI scored from evidence
+7. **Callable advisor YAML** — ready to invoke from any /arka command
 
-Persona profile in Obsidian with full behavioral DNA and mental model catalog
+## Source Inventory Format
+
+Every persona source is tagged:
+
+```yaml
+sources:
+  - id: source-001
+    url: <url or local path>
+    type: talk | interview | written | social | podcast | book
+    date: YYYY-MM-DD
+    weight: high | medium | low   # based on persona's intent + content depth
+    summary: <one-line description>
+```
+
+Minimum source diversity for a credible persona: 5 sources across at least 3 types, spanning at least 6 months of date range.
+
+## Belief Inventory (cited extraction)
+
+A belief is a claim the persona has made consistently across multiple sources. Format:
+
+```yaml
+beliefs:
+  - statement: "<the belief in plain language>"
+    strength: load-bearing | supporting | aspirational
+    citations:
+      - source: source-003
+        quote: "<verbatim>"
+      - source: source-007
+        quote: "<verbatim>"
+    counter-position: <what this belief rejects, if explicit>
+```
+
+Beliefs without 2+ citations are dropped — single-mention statements are not strong enough to be a persona trait. Load-bearing beliefs (the ones the persona cannot abandon without becoming someone else) require 4+ citations.
+
+## Voice Pattern Extraction
+
+The voice signature has 4 layers:
+
+### 1. Lexical signature (vocabulary)
+- **Proprietary terms** — words the persona uses unusually often or coined
+- **Forbidden words** — words the persona never uses (often due to belief)
+- **N-gram patterns** — 2-3 word phrases that recur in characteristic positions
+
+### 2. Syntactic signature (sentence structure)
+- Average sentence length
+- Clause complexity (simple / compound / complex preferred)
+- Sentence-opening patterns (3-5 typical openers)
+- Sentence-closing patterns (3-5 typical closes)
+
+### 3. Rhetorical signature (argumentation pattern)
+- Preferred argument structure (claim-evidence-warrant vs Socratic vs analogy-driven vs storytelling)
+- Use of qualifiers (hedged vs declarative)
+- Use of self-disclosure (front-loaded / withheld / never)
+
+### 4. Tonal signature (emotional register)
+- Default warmth (cold / neutral / warm)
+- Default certainty (hedged / measured / declarative)
+- Humor pattern (none / wry / direct / self-deprecating)
+- Confrontation pattern (avoided / oblique / direct)
+
+Each layer must be derivable from at least 3 source examples.
+
+## Expertise Map
+
+Three concentric domains:
+
+```yaml
+expertise:
+  deep:
+    - domain: <area>
+      evidence: <which sources prove deep expertise>
+      sub-domains: [<list>]
+  surface:
+    - domain: <area>
+      evidence: <touched briefly in sources>
+  no-go:
+    - domain: <area>
+      reason: <why this persona refuses to opine here>
+      citation: <source where they explicitly declined>
+```
+
+A persona that opines on everything has no persona — the no-go list is as important as the deep list.
+
+## Decision Patterns
+
+Extract characteristic decision-making patterns with cited examples:
+
+```yaml
+decision_patterns:
+  - pattern: "<named pattern>"
+    description: <plain language>
+    example:
+      context: <the decision>
+      reasoning: <how they thought about it>
+      source: <cited>
+    inverse: <what they would NOT do in that context>
+```
+
+Examples of named patterns: "Optimises for irreversibility avoidance", "Prefers small-bet portfolio over big-bet concentration", "Refuses to publish until idea has been steel-manned".
+
+## Behavioral DNA (4-framework score)
+
+Score each framework based on source evidence, NOT speculation. Each scored dimension cites the source.
+
+### DISC profile
+- D (Dominance): 0-100 with source evidence
+- I (Influence): 0-100 with source evidence
+- S (Steadiness): 0-100 with source evidence
+- C (Conscientiousness): 0-100 with source evidence
+- Type: Dominant pair (D+C, S+C, I+D, etc.)
+
+### Enneagram
+- Type: 1-9
+- Wing: w-X
+- Health level evidence: 1-9 from sources
+- Core motivation cited: <source>
+- Core fear cited: <source>
+
+### Big Five (OCEAN)
+- Openness: 0-100
+- Conscientiousness: 0-100
+- Extraversion: 0-100
+- Agreeableness: 0-100
+- Neuroticism: 0-100
+Each with 2+ source citations.
+
+### MBTI
+- 4 letters with cited preference per letter
+- Cognitive function stack derivation
+
+## Callable Advisor YAML (production output)
+
+The final deliverable is a YAML that any /arka command can load:
+
+```yaml
+id: <persona-slug>
+name: <display name>
+canonical_reference: <real human, redacted if confidentiality requires>
+expertise_domains_deep: [<list>]
+expertise_domains_surface: [<list>]
+expertise_domains_nogo: [<list>]
+voice_pattern:
+  lexical_signature_path: <obsidian link>
+  syntactic_signature_path: <obsidian link>
+  rhetorical_pattern: <name>
+  tonal_signature: <warmth + certainty + humor + confrontation>
+beliefs:
+  load_bearing: [<list with source IDs>]
+  supporting: [<list with source IDs>]
+decision_patterns: [<named patterns>]
+behavioral_dna:
+  disc: { D: X, I: X, S: X, C: X }
+  enneagram: { type: X, wing: X }
+  big_five: { O: X, C: X, E: X, A: X, N: X }
+  mbti: <4 letters>
+invocation_examples:
+  - context: <when to call this persona>
+    expected_output: <what this persona would produce>
+```
+
+## Common Failure Modes
+
+1. **Persona without no-go domains** — a persona that opines on everything reads as a generic AI, not a specific human
+2. **Beliefs without citations** — single-source claims are not beliefs, they're moments. Drop them
+3. **Voice signature copied from prose, not patterns** — describing the voice ("authoritative", "direct") doesn't replicate it. Extract structural N-grams
+4. **DNA scored from speculation** — every behavioral score must trace to source evidence
+5. **Persona that's actually a category** — "Tech founder" is not a persona, "Paul Graham" is. Personas need a named reference human
+
+## Output → Obsidian: `WizardingCode/Personas/<name>/`
+
+Delivers: source inventory + belief inventory (cited) + voice pattern (4 layers) + expertise map (deep/surface/no-go) + decision patterns + 4-framework DNA + callable advisor YAML + invocation examples.

package/departments/kb/workflows/persona.yaml

@@ -0,0 +1,131 @@
+id: kb-persona
+name: Persona Build
+description: Build or refine a persona from the KB — extract beliefs, voice, expertise, decision patterns; produce a callable advisor profile
+department: kb
+tier: enterprise
+command: "/kb persona"
+requires_branch: false
+requires_spec: false
+quality_gate_required: true
+
+phases:
+  - id: brief
+    name: Persona Brief
+    description: Define who the persona represents — real human reference, target use cases, advisor context
+    agents:
+      - agent_id: kb-curator-clara
+        role: Frame the persona — named reference human, role, decision domains, advisor use cases
+    gate:
+      type: user_approval
+      description: User confirms persona scope and reference
+
+  - id: source-gathering
+    name: Source Gathering
+    description: Collect all KB sources for the persona — talks, interviews, writings, transcripts, social posts
+    agents:
+      - agent_id: kb-curator-clara
+        role: KB source enumeration with date range and source-type tagging
+    gate:
+      type: user_approval
+      description: User approves source set
+    outputs:
+      - type: document
+        format: markdown
+        obsidian_path: "WizardingCode/Personas/Sources/"
+        description: Source inventory per persona
+
+  - id: belief-extraction
+    name: Belief & Frame Extraction
+    description: Extract the persona's core beliefs, mental models, and recurring frames from sources
+    agents:
+      - agent_id: kb-curator-clara
+        role: Belief extraction with source citation per belief
+      - agent_id: research-analyst-pedro
+        role: Pattern detection across sources — what recurs, what's load-bearing
+        parallel: true
+    gate:
+      type: user_approval
+      description: User approves belief inventory
+
+  - id: voice-pattern
+    name: Voice & Vocabulary Pattern
+    description: Map the persona's voice — sentence rhythm, vocabulary signature, characteristic phrases, no-go words
+    agents:
+      - agent_id: kb-curator-clara
+        role: Voice analysis with N-gram patterns + recurring phrases + structural signatures
+      - agent_id: copy-director-eduardo
+        role: Voice replication test — would a current reader recognise this voice?
+        parallel: true
+    gate:
+      type: user_approval
+      description: User approves voice pattern
+
+  - id: expertise-domains
+    name: Expertise Domains
+    description: Map what the persona knows deeply vs surface-level vs explicitly avoids
+    agents:
+      - agent_id: kb-curator-clara
+        role: Expertise mapping — 3-5 deep domains, 5-10 surface domains, explicit no-go domains
+    gate:
+      type: user_approval
+      description: User approves expertise map
+
+  - id: decision-patterns
+    name: Decision Patterns
+    description: Extract characteristic decision patterns — how the persona evaluates trade-offs, what evidence they accept, what they refuse
+    agents:
+      - agent_id: kb-curator-clara
+        role: Decision pattern extraction with example decisions cited from sources
+    gate:
+      type: user_approval
+      description: User approves decision pattern map
+
+  - id: advisor-profile
+    name: Callable Advisor Profile
+    description: Compile the persona into a callable advisor YAML — voice + beliefs + expertise + decision patterns + bio
+    agents:
+      - agent_id: kb-curator-clara
+        role: Advisor profile YAML compilation
+    gate:
+      type: user_approval
+      description: User approves callable advisor profile
+
+  - id: self-critique
+    name: Self-Critique
+    description: Stress-test the persona — would the reference human recognise themselves? Are beliefs cited? Are no-go domains enforceable?
+    agents:
+      - agent_id: kb-curator-clara
+        role: Self-test — generate 3 sample advisor outputs and verify voice + beliefs
+    gate:
+      type: auto
+
+  - id: quality-gate
+    name: Quality Gate
+    model_override: opus
+    description: Mandatory quality review
+    agents:
+      - agent_id: cqo-marta
+        role: Orchestrate quality review
+      - agent_id: copy-director-eduardo
+        role: Voice fidelity, vocabulary signature accuracy, no-clichés
+        parallel: true
+      - agent_id: tech-director-francisca
+        role: Source citation integrity, expertise domain boundaries, decision pattern falsifiability
+        parallel: true
+    gate:
+      type: quality_gate
+      required_verdict: APPROVED
+
+  - id: delivery
+    name: Persona Package Delivery
+    description: Compile the persona package — sources + beliefs + voice + expertise + decisions + advisor YAML
+    agents:
+      - agent_id: kb-curator-clara
+        role: Full persona package + advisor invocation guide
+    gate:
+      type: auto
+    outputs:
+      - type: document
+        format: markdown
+        obsidian_path: "WizardingCode/Personas/"
+        description: Complete persona — sources + beliefs (cited) + voice pattern + expertise map + decision patterns + advisor YAML + invocation guide

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "arkaos",
-  "version": "2.56.0",
+  "version": "2.57.0",
   "description": "The Operating System for AI Agent Teams",
   "type": "module",
   "bin": {

package/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "arkaos-core"
-version = "2.56.0"
+version = "2.57.0"
 description = "Core engine for ArkaOS — The Operating System for AI Agent Teams"
 readme = "README.md"
 license = {text = "MIT"}