Game Audio Feedback Accessibility

Skill Registry

Skill Metadata

• Name: game-audio-feedback-accessibility
• Version: 1.0.0
• Schema Version: 1.0
• Domain: Audio-Based Game Accessibility & Sonification
• Type: Harness Orchestrator
• Language Support: English (en), Vietnamese (vi)
• Runtime: Claude Code with MCP tools

Skill Architecture

game-audio-feedback-accessibility (Main Harness)
├── Hook System (Lifecycle & Event Management)
├── Tool Registry (Dynamic Tool Invocation)
├── Skill Orchestrator (Sequential Sub-Skill Execution)
├── Quality Gate Engine (Verification & Auto-Fix)
├── Knowledge Pipeline (Self-Improving Crawl)
└── Graceful Degradation (Level-Based Fallback)

Sub-Skills Registration

| Sub-Skill | Purpose | Input Schema | Output Schema | |-----------|---------|--------------|---------------| | sub-gather-requirements | Clarify analysis parameters | {raw_query, available_inputs} | {object, scope, timeframe, inputs, audience, language} | | sub-evidence-collector | Fetch authoritative data | {requirements} | {current_data, authoritative_docs, recent_news} | | sub-core-analysis | Design audio-feedback system | {game, required_info, audience, language} | {cue_mapping, sonification, redundancy, customization, load, scenarios} | | sub-knowledge-updater | Query knowledge base | {topic_keywords} | {citations: [{tier, source, relevance}]} | | sub-advisor | Synthesize risk-disclosed conclusion | {analysis, evidence, knowledge} | {verdict, scenarios, risks, evidence_chain, remediation, disclosure} |

Tool Definitions

WebResearchTool

• Purpose: Fetch real-time domain data from authoritative sources
• Schema:

``json { "type": "object", "properties": { "query": {"type": "string"}, "sources": {"type": "array", "items": {"type": "string"}}, "tier": {"type": "string", "enum": ["authoritative", "academic", "industry", "general"]} } } ``

KnowledgeQueryTool

• Purpose: Query SECOND-KNOWLEDGE-BRAIN.md for academic evidence
• Schema:

``json { "type": "object", "properties": { "keywords": {"type": "array", "items": {"type": "string"}}, "min_tier": {"type": "integer", "minimum": 1, "maximum": 4} } } ``

AudioDesignTool

• Purpose: Apply audio accessibility design patterns
• Schema:

``json { "type": "object", "properties": { "game_elements": {"type": "array"}, "audience": {"type": "string", "enum": ["blind", "low-vision", "mixed"]}, "redundancy_level": {"type": "string", "enum": ["minimal", "standard", "high"]} } } ``

Hook System

Lifecycle Hooks

• pre-analysis: Before sub-skills execute
• post-evidence: After evidence collection
• post-analysis: After core analysis
• pre-quality-gate: Before final gate verification
• post-output: After final output generation

Event Emission

• AnalysisStarted: {timestamp, query_id, requirements}
• EvidenceCollected: {timestamp, sources_count, data_points}
• **AnalysisComplete: {timestamp, verdict, confidence}`
• QualityGatePassed: {timestamp, gate_id, result}
• AnalysisFailed: {timestamp, error, degradation_level}

State Synchronization

• AnalysisState: Shared across sub-skills
• EvidenceCache: Persists between sessions
• KnowledgeIndex: References SECOND-KNOWLEDGE-BRAIN.md

Quality Gate Schema

Universal Gates (U1-U6)

• U1: Evidence Quantity (≥3 sources, ≥1 academic)
• U2: Risk Disclosure (before recommendation)
• U3: Evidence Hierarchy (tier labels)
• U4: Language Matching
• U5: Template Compliance
• U6: Claim Traceability

Domain Gates (G1-G4)

• G1: Required Info Mapping (all info → audio cues)
• G2: Redundancy (audio-haptic + screen-reader)
• G3: Customization (volume/pan/speed controls)
• G4: Cognitive Load Management

Configuration Management

See config/default.json for runtime configuration:

• LLM parameters
• Feature flags
• Source priorities
• Timeout settings
• Degradation thresholds

Error Handling

Error Types

• SourceTimeout: No response within 30s
• InvalidInput: Schema mismatch
• MissingInput: Required field absent
• StaleData: Timestamp exceeds threshold
• KnowledgeBaseMiss: No matches found
• ConflictingActions: Mutually exclusive operations
• EnvelopeUnavailable: No setpoint available
• ObjectAmbiguous: Classification unclear

Recovery Strategies

• Retry with alternate source (max 3 attempts)
• Ask user for confirmation
• Proceed with available data + flag
• Flag stale, request refresh
• WebSearch gap-fill + queue crawl
• Apply stated precedence
• Use genus/category fallback
• Ask user to confirm

Output Template

# Audio Feedback System Design for Visually-Impaired Accessibility — Report
**Date:** {date} | **Analyst:** {version} | **Language:** {language} | **Domain:** {domain}

## Executive Summary
{summary}

## Inputs & Scope
{requirements}

## Evidence Collected
{evidence_with_sources_and_tiers}

## Analysis / Scorecard
{analysis_results}

## Action / Control Plan
{actions}

## Academic & Research Evidence
{academic_citations}

## ⚠️ Disclosure / Limitations
{disclosure}

## Recommendation / Conclusion
{verdict_with_scenarios_and_risks}

## Post-Execution Gate Checklist
{gate_checklist}

Execution Flow

1. Pre-Flight: Language Detection
2. Hook: pre-analysis
3. Invoke sub-gather-requirements
4. Invoke sub-evidence-collector
5. Hook: post-evidence
6. Invoke sub-core-analysis
7. Hook: post-analysis
8. Invoke sub-knowledge-updater
9. Invoke sub-advisor
10. Hook: pre-quality-gate
11. Quality Gate Verification (U1-U6, G1-G4)
12. Hook: post-output
13. Deliver Final Report

Graceful Degradation Levels

| Level | Condition | Behavior | |-------|-----------|----------| | 0 | All sources reachable | Full analysis | | 1 | Some sources fail | Secondary sources + flag | | 2 | Most sources fail | Knowledge base only + historical flag | | 3 | Input missing/stale | Available variables + DATA UNAVAILABLE | | 4 | All sources + KB fail | Emit DATA UNAVAILABLE notice |

Self-Improving Knowledge Pipeline

• Crawl Sources: ArXiv, Semantic Scholar, RSS feeds
• Deduplication: SHA256 of DOI/URL
• Scoring: Recency (40%) + Keyword Relevance (40%) + Citation Count (20%)
• Schedule: Weekly academic (Mon 08:00) + Daily news (07:00)
• Tool: tools/knowledge_updater.py

Version History

• 1.0.0 (2026-07-15): Production-grade release with enhanced architecture
• 2.0.0 (2026-07-28): Bulletproof production upgrade with router, monitoring, API, and optimization

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: dungnotnull
• Source: dungnotnull/game-audio-feedback-accessibility-agent-skill
• License: MIT

Install and usage instructions live in the source repository linked above.