hushh-labs

morphy-ax

Use when changing Morphy AX shared agent-experience state, semantic assessment presentation, performance budgets, or cross-modal compatibility.

hushh-labs 40 87 Updated 3mo ago

Resources

1
GitHub

Install

npx skillscat add hushh-labs/hushh-research/morphy-ax

Install via the SkillsCat registry.

SKILL.md

Morphy AX

Purpose and Trigger

  • Primary scope: morphy-agent-experience
  • Trigger on Morphy AX snapshots, active interaction layers, presentation derivation, semantic-assessment validation, compatibility projections, or AX performance budgets.
  • Avoid overlap with frontend-design-system, one-voice-governance, and backend-agents-operons.

Coverage and Ownership

  • Role: spoke
  • Owner family: frontend

Owned repo surfaces:

  1. hushh-webapp/lib/morphy-ax
  2. docs/reference/quality/morphy-agent-experience.md

Non-owned surfaces:

  1. hushh-webapp/lib/voice
  2. consent-protocol/hushh_mcp/agents
  3. hushh-webapp/lib/morphy-ux

Do Use

  1. Shared, redacted agent-experience snapshot, active-layer composition, and presentation contracts.
  2. Intelligence-assessment validation and compatibility/non-regression budgets.
  3. Cross-modal AX behavior that reuses the existing runtime owner and action gateway.

Do Not Use

  1. Voice action authoring, generated gateway changes, or ADK orchestration ownership.
  2. Visual primitive work owned by Morphy UX.
  3. A second React provider, store, router, action registry, DOM observer, or model-facing route tool.

Read First

  1. docs/reference/quality/design.md
  2. docs/reference/quality/morphy-agent-experience.md
  3. docs/reference/one/one-voice-runtime-architecture.md
  4. docs/reference/quality/frontend-ui-architecture-map.md
  5. docs/reference/kai/kai-action-gateway-vnext.md

Workflow

  1. Verify the existing runtime, voice FSM, route playbook, and generated action authority before changing AX.
  2. Keep Morphy AX pure, redacted, memoizable, and hosted by AgentRuntimeStateProvider.
  3. Let intelligence assess meaning; deterministic code may validate, reject, normalize, and enforce authority only.
  4. Compose authored route, chrome, and interaction_layer publishers; expose only the top effective inventory and remove stale owner leases on route change.
  5. Require exact generated dismiss/cancel actions, focus-restored settlement, and declared Agent continuity; never synthesize click or Escape behavior.
  6. Treat trusted_activation_required browser actions as exact Agent Bar continuations over mounted handlers, not redirect or popup-broker fallbacks.
  7. Preserve OneVoiceContextSnapshot through an explicit compatibility projection while migration is active.
  8. Benchmark the same fixtures flag-off and flag-on; block wrong-layer actions, unsafe fallbacks, or budget regressions.
  9. Route voice, backend-agent, design-system, docs, or quality changes to their owning skills.

Handoff Rules

  1. Broad frontend work routes to frontend.
  2. Voice/action contracts route to one-voice-governance.
  3. ADK or product-agent changes route to backend-agents-operons.
  4. Visual primitives route to frontend-design-system; verification policy routes to quality-contracts.

Required Checks

cd hushh-webapp && npm run verify:morphy-ax
cd hushh-webapp && npm run typecheck
cd hushh-webapp && npm run verify:voice-gateway
cd hushh-webapp && npm run verify:design-system
cd consent-protocol && python3 -m pytest tests/test_onboarding_goal_agent.py tests/test_one_adk_agent_tree.py -q
./bin/hushh docs verify