Update STATE.md after Domain 20 build

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: GarethManning

STATE.md

@@ -1,62 +1,29 @@
-# STATE.md — education-agent-skills
+# State — Education Agent Skills Library
 
-**Last updated:** 2026-05-19 (Session 7)
+## Last updated: 2026-05-27
 
-## What was done
+## What was done this session
 
-Added 5 H3Uni skills across two domains:
+Built and shipped Domain 20: Student-Facing Learning Skills — 13 new skills in a new domain `student-learning`.
 
-**systems-thinking:**
-1. **systems-thinking/hexagon-complexity-mapper** — Guides students through hexagon mapping: factors placed on tiles where adjacency signals a claimed relationship. Critical principle: adjacency IS the meaning-making gesture, not sorting or categorisation. Evidence: Hodgson (1992), Meadows (2008), Senge (1990), H3Uni guides. evidence_strength: emerging.
+All 13 skills: retrieve-first-gate, explain-first-interrogator, progressive-hint-ladder, confidence-calibration-check, stuck-and-error-diagnosis-coach, ai-claim-checker, transfer-bridge, teach-back-evaluator, productive-failure-protocol, srl-session-wrapper, unassisted-evidence-checkpoint, weekly-agency-review, fading-manager.
 
-**original-frameworks:**
-2. **original-frameworks/scoping-for-transformative-learning-inquiry** — Defines inquiry scope before any mapping begins. Output: scoping statement in format "We are exploring [what] from the perspective of [who] so that we can [purpose]." Evidence: H3Uni Scoping method, Fazey et al. (2018), Rajagopalan & Midgley (2015). evidence_strength: emerging.
-3. **original-frameworks/three-horizons-learning-transition-mapper** — Maps H1 (current pattern under strain) → H3 (preferred future) → H2 (innovations sorted as H2+ or H2-). Critical sequence enforced: never H1 → H2 → H3. Evidence: Curry & Hodgson (2008), Sharpe et al. (2016), Sharpe & Hodgson (2019), H3Uni guides. evidence_strength: emerging.
-4. **original-frameworks/dilemma-navigation-for-education-design** — Navigates genuine dilemmas (both/and integration, not winner-picking). Critical distinction: dilemmas have legitimate value on both sides; problems have solutions. Evidence: H3Uni Dilemma Navigation, Johnson (1992), Hampden-Turner (1990). evidence_strength: emerging.
-5. **original-frameworks/multi-perspective-decision-wheel** — Examines decisions through 4–6 perspectives (each with: sees, misses, asks, contributes), names tensions, synthesises insights, and produces a next responsible step. Evidence: H3Uni Wheel of Wisdom, Midgley (2000), Rajagopalan & Midgley (2015). evidence_strength: emerging.
+## What was verified
 
-Also:
-- README badge updated: 147 → 152 skills
-- README domain table updated: rows 4 (4→5), 7 (15→13), 9 (9→10), 15 (11→17), 18 (5→8); descriptions updated for original-frameworks and systems-thinking rows
-- README MCP tool count updated: 149 → 156 tools, 145 → 152 prompts
-- CLAUDE.md updated: 147 → 152 skills
-- llms.txt updated: 131 → 152 skills, 17 → 19 domains; all 19 domain rows with correct counts
+- All 20 Playwright tests pass (was 18 before, 2 new skills-library tests were already in suite)
+- registry.json regenerated: 165 skills, 20 domains
+- mcp-server/src/skills.json rebundled: 165 skills
+- Committed and pushed: d84b8f4
 
-## What was verified (Session 7)
+## Current library state
 
-- `python3 scripts/generate-registry.py`: 152 skills, 19 domains
-- `python3 scripts/validate-skills.py`: 152 skills, 0 errors, 4 pre-existing warnings (line-length on unrelated skills)
-- `python3 scripts/validate-registry.py`: 152 skills, 19 domains — valid
-- `npx playwright test`: 20/20 pass
-- `cd mcp-server && npm run bundle-skills`: 152 skills bundled
-- `cd mcp-server && npm run build && npm test`: 16/16 pass
-
-## Session 6 summary (2026-05-18)
-
-Added 2 orchestrator skills (assessment-design-orchestrator, inclusive-design-orchestrator). Registry at 147 skills. Committed: b88d3b8.
-
-## Chain targets verified
-
-Assessment Design Orchestrator:
-- formative-assessment-technique-selector ✓
-- assessment-validity-checker ✓
-- gap-analysis-from-student-work ✓
-- differentiation-adapter ✓
-- self-regulation-scaffold-generator ✓
-- inclusive-design-orchestrator ✓ (created this session)
-- udl-barrier-anticipator ✓
-
-Inclusive Design Orchestrator:
-- udl-lesson-auditor ✓
-- udl-options-designer ✓
-- udl-barrier-anticipator ✓
-- differentiation-adapter ✓
-- assessment-design-orchestrator ✓ (created this session)
-- language-demand-analyser ✓
-- scaffolded-task-modifier ✓
+- 165 skills across 20 domains
+- Domain 20 is new and different from Domains 1–19: student-facing live interaction patterns, not educator-facing artefact generation
+- Domain 20 introduces `evidence_captured` YAML schema extension for structured learning evidence
+- `generate-registry.py` updated with "student-learning" in DOMAIN_LABELS
 
 ## What's next
 
-- Systems-thinking orchestrator: systems-thinking domain now has 8 skills (iceberg, aspirational iceberg, hexagon complexity mapper, leverage-response design, mental model mapper, agency circles, ladder of inference, wellbeing impact mapper). A composite orchestrator linking these into a full inquiry arc is the logical next step (separate from the existing compassionate-systems-awareness-orchestrator which covers a subset).
-- Perspective-taking designer (questioning-discussion) has natural cross-domain chains with historical-thinking — worth formalising.
-- The assessment-design-orchestrator and inclusive-design-orchestrator now cross-chain with each other, making a combined "accessible assessment design" flow possible as a future composite skill.
+- Downstream: Kaku's Tutor Agent and Hippo can now reference Domain 20 skills
+- Consider building a cross-domain orchestrator skill that chains Domain 20 skills (e.g., 20-01 → 20-04 → 20-11 sequence)
+- QA evaluator session (separate Claude invocation) recommended before production use of Domain 20 skills