Game Livestream Multilingual Captioning

SKILL.md — Game Livestream Multilingual Captioning Registry

Overview

This document serves as the central registry for the game-livestream-multilingual-captioning skill system. It defines the complete architecture for skill registration, resolution, execution, and validation within the Claude Code harness environment.

Skill Identity

skill_id: game-livestream-multilingual-captioning
version: 1.0.0
domain: Live Stream Multilingual Captioning & Speech Translation
namespace: glmc
entry_point: /game-livestream-multilingual-captioning
cli_entry: glmc

Architecture Pattern

The skill follows a Modular Sub-Skill Pattern with:

Main Harness (main.md)
    │
    ├──► Pre-Flight: Language Detection
    │
    ├──► Sub-Skill 1: Requirements Gathering
    ├──► Sub-Skill 2: Evidence Collection
    ├──► Sub-Skill 3: Core Analysis
    ├──► Sub-Skill 4: Knowledge Update
    ├──► Sub-Skill 5: Advisory Synthesis
    │
    └──► Quality Gate Enforcement (10 gates)

Skill Registration

Frontmatter Contract

Every skill file MUST include YAML frontmatter:

---
name: {skill-or-subskill-name}
description: {one-line summary explaining when to trigger}
---

Registration Schema

{
  "skill": {
    "name": "string",
    "description": "string",
    "version": "string",
    "type": "main|sub",
    "namespace": "string",
    "entry_points": {
      "cli": "string",
      "api": "string",
      "skill_path": "string"
    },
    "dependencies": {
      "tools": ["string"],
      "sub_skills": ["string"],
      "external": ["string"]
    },
    "metadata": {
      "author": "string",
      "created": "ISO8601",
      "updated": "ISO8601",
      "tags": ["string"]
    }
  }
}

Sub-Skill Registry

| Sub-Skill | Purpose | Input Schema | Output Schema | Dependencies | |-----------|---------|--------------|---------------|--------------| | sub-gather-requirements | Intake specialist | RequirementsInput | RequirementsOutput | Conversation | | sub-evidence-collector | Data librarian | EvidenceInput | EvidenceOutput | WebSearch, WebFetch, Read | | sub-core-analysis | Pipeline engineer | AnalysisInput | AnalysisOutput | Read, WebFetch, Reasoning | | sub-knowledge-updater | Research librarian | KnowledgeInput | KnowledgeOutput | Read, WebSearch | | sub-advisor | Senior advisor | AdvisoryInput | AdvisoryOutput | Reasoning, Skill |

Input/Output JSON Schemas

RequirementsInput (sub-gather-requirements)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "RequirementsInput",
  "type": "object",
  "properties": {
    "raw_message": {
      "type": "string",
      "description": "Raw user input message"
    },
    "context": {
      "type": "object",
      "properties": {
        "language": {"type": "string", "enum": ["vi", "en", "zh", "ja", "ko", "auto"]},
        "previous_analysis": {"type": "object"},
        "user_preferences": {"type": "object"}
      }
    },
    "provided_materials": {
      "type": "array",
      "items": {"type": "string"},
      "description": "File paths or URLs provided by user"
    }
  },
  "required": ["raw_message"]
}

RequirementsOutput (sub-gather-requirements)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "RequirementsOutput",
  "type": "object",
  "properties": {
    "object": {
      "type": "string",
      "description": "Primary object of analysis"
    },
    "platform": {
      "type": "string",
      "enum": ["twitch", "youtube", "facebook", "kick", "obs", "custom", "unknown"]
    },
    "source_languages": {
      "type": "array",
      "items": {"type": "string"},
      "description": "Source language(s) for transcription"
    },
    "target_languages": {
      "type": "array",
      "items": {"type": "string"},
      "description": "Target language(s) for translation"
    },
    "hardware_tier": {
      "type": "integer",
      "minimum": 1,
      "maximum": 5,
      "description": "Hardware capability tier (1-5)"
    },
    "game_genre": {
      "type": "string",
      "enum": ["fps", "moba", "rpg", "strategy", "sports", "chatting", "other", "unknown"]
    },
    "scope": {
      "type": "string",
      "enum": ["full_pipeline", "asr_only", "mt_only", "comparison", "feasibility"]
    },
    "latency_requirement_ms": {
      "type": "integer",
      "description": "Target end-to-end latency in milliseconds"
    },
    "analysis_type": {
      "type": "string",
      "enum": ["production", "feasibility", "comparison", "troubleshooting", "educational"]
    },
    "red_flags": {
      "type": "array",
      "items": {"type": "string"},
      "description": "Detected red flags or constraints"
    },
    "language": {
      "type": "string",
      "enum": ["vi", "en"],
      "description": "Output language preference"
    }
  },
  "required": ["object", "platform", "source_languages", "target_languages", "hardware_tier", "language"]
}

EvidenceInput (sub-evidence-collector)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "EvidenceInput",
  "type": "object",
  "properties": {
    "requirements": {
      "$ref": "#/definitions/RequirementsOutput"
    },
    "evidence_tiers": {
      "type": "object",
      "description": "Source priority configuration"
    }
  },
  "required": ["requirements"]
}

EvidenceOutput (sub-evidence-collector)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "EvidenceOutput",
  "type": "object",
  "properties": {
    "current_data": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "source": {"type": "string"},
          "content": {"type": "string"},
          "access_date": {"type": "string", "format": "date-time"},
          "tier": {"type": "string", "enum": ["P0", "P1", "P2", "P3", "FALLBACK"]}
        }
      }
    },
    "authoritative_docs": {
      "type": "array",
      "items": {"$ref": "#/definitions/EvidenceItem"}
    },
    "recent_news": {
      "type": "array",
      "items": {"$ref": "#/definitions/EvidenceItem"}
    },
    "reference_benchmarks": {
      "type": "object",
      "properties": {
        "asr_models": {"type": "object"},
        "mt_engines": {"type": "object"},
        "voice_separation": {"type": "object"}
      }
    },
    "degradation_level": {
      "type": "integer",
      "minimum": 0,
      "maximum": 4
    }
  }
}

AnalysisInput (sub-core-analysis)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "AnalysisInput",
  "type": "object",
  "properties": {
    "requirements": {
      "$ref": "#/definitions/RequirementsOutput"
    },
    "evidence": {
      "$ref": "#/definitions/EvidenceOutput"
    }
  },
  "required": ["requirements", "evidence"]
}

AnalysisOutput (sub-core-analysis)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "AnalysisOutput",
  "type": "object",
  "properties": {
    "pipeline_design": {
      "type": "object",
      "properties": {
        "voice_separation": {"type": "object"},
        "asr_config": {"type": "object"},
        "mt_pipeline": {"type": "object"},
        "caption_encoding": {"type": "object"},
        "latency_budget": {"type": "object"},
        "quality_metrics": {"type": "object"}
      }
    },
    "scenarios": {
      "type": "array",
      "items": {"$ref": "#/definitions/Scenario"}
    },
    "scorecard": {
      "$ref": "#/definitions/Scorecard"
    },
    "failure_modes": {
      "type": "array",
      "items": {"$ref": "#/definitions/FailureMode"}
    }
  }
}

AdvisoryInput (sub-advisor)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "AdvisoryInput",
  "type": "object",
  "properties": {
    "analysis": {
      "$ref": "#/definitions/AnalysisOutput"
    },
    "evidence": {
      "$ref": "#/definitions/EvidenceOutput"
    },
    "knowledge": {
      "type": "array",
      "items": {"$ref": "#/definitions/KnowledgeEntry"}
    }
  },
  "required": ["analysis", "evidence"]
}

AdvisoryOutput (sub-advisor)

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "AdvisoryOutput",
  "type": "object",
  "properties": {
    "verdict": {
      "type": "string",
      "enum": ["Production-Ready Pipeline", "Conditional", "High Latency Risk", "Inconclusive"]
    },
    "confidence": {
      "type": "string",
      "enum": ["HIGH", "MEDIUM", "LOW"]
    },
    "scenarios": {
      "type": "array",
      "items": {"$ref": "#/definitions/Scenario"}
    },
    "key_risks": {
      "type": "array",
      "items": {"$ref": "#/definitions/Risk"}
    },
    "evidence_chain": {
      "type": "array",
      "items": {"type": "object"}
    },
    "remediation_priorities": {
      "type": "array",
      "items": {"$ref": "#/definitions/Remediation"}
    },
    "deployment_checklist": {
      "type": "array",
      "items": {"type": "string"}
    },
    "mandatory_disclosure": {
      "type": "string"
    }
  },
  "required": ["verdict", "confidence", "mandatory_disclosure"]
}

Quality Gate Schema

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "QualityGate",
  "type": "object",
  "properties": {
    "gate_id": {
      "type": "string",
      "pattern": "^[UG][0-9]$"
    },
    "description": {"type": "string"},
    "check_type": {
      "type": "string",
      "enum": ["automated", "manual", "hybrid"]
    },
    "auto_fix_available": {"type": "boolean"},
    "auto_fix_procedure": {"type": "string"},
    "enforcement_level": {
      "type": "string",
      "enum": ["block", "warn", "log"]
    },
    "retry_limit": {"type": "integer"},
    "status": {
      "type": "string",
      "enum": ["pending", "passed", "failed", "degraded"]
    }
  },
  "required": ["gate_id", "description", "check_type", "enforcement_level"]
}

Execution Flow Contract

Main Harness Execution

1. PRE-FLIGHT: Language Detection
   ├─ Input: raw_message
   ├─ Process: Detect language from message content
   └─ Output: language_preference (vi/en/zh/ja/ko)

2. STEP 1: Requirements Gathering
   ├─ Input: raw_message + context
   ├─ Process: Extract structured requirements
   └─ Output: RequirementsOutput

3. STEP 2: Evidence Collection
   ├─ Input: RequirementsOutput
   ├─ Process: Fetch authoritative data by tier
   └─ Output: EvidenceOutput

4. STEP 3: Core Analysis
   ├─ Input: RequirementsOutput + EvidenceOutput
   ├─ Process: Design full pipeline
   └─ Output: AnalysisOutput

5. STEP 4: Knowledge Update
   ├─ Input: AnalysisOutput + topic_keywords
   ├─ Process: Query KB, surface citations
   └─ Output: KnowledgeOutput

6. STEP 5: Advisory Synthesis
   ├─ Input: AnalysisOutput + EvidenceOutput + KnowledgeOutput
   ├─ Process: Synthesize risk-disclosed conclusion
   └─ Output: AdvisoryOutput

7. QUALITY GATE: 10-Gate Check
   ├─ Process: Validate all gates
   ├─ Auto-fix: Apply fixes for failed gates
   ├─ Retry: Max 2 retries per gate
   └─ Output: Final report or degradation notice

Error Handling Contract

{
  "error_handling": {
    "strategy": "graceful_degradation",
    "levels": [0, 1, 2, 3, 4],
    "fallback_chain": [
      "primary_sources",
      "secondary_sources",
      "knowledge_base",
      "cached_results",
      "limitation_notice"
    ],
    "retry_policy": {
      "max_retries": 2,
      "backoff": "exponential",
      "retryable_errors": ["timeout", "rate_limit", "temporary_failure"]
    }
  }
}

Validation Schema

Skill Health Check

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "SkillHealthCheck",
  "type": "object",
  "properties": {
    "skill_name": {"type": "string"},
    "version": {"type": "string"},
    "status": {
      "type": "string",
      "enum": ["healthy", "degraded", "unhealthy"]
    },
    "sub_skills_status": {
      "type": "object",
      "additionalProperties": {
        "type": "string",
        "enum": ["available", "degraded", "unavailable"]
      }
    },
    "dependencies": {
      "type": "object",
      "properties": {
        "tools": {"type": "object"},
        "external_services": {"type": "object"}
      }
    },
    "quality_gates": {
      "type": "object",
      "properties": {
        "total": {"type": "integer"},
        "passed": {"type": "integer"},
        "failed": {"type": "integer"},
        "degraded": {"type": "integer"}
      }
    }
  }
}

CLI Integration

Command Registration

# Primary entry point
glmc analyze "query"           # Full pipeline execution
glmc analyze "query" --json    # Machine-readable output
glmc analyze "query" -l vi     # Vietnamese output

# Utility commands
glmc validate                  # Run 8-File Contract validation
glmc test                      # Run test scenarios
glmc update-knowledge          # Execute knowledge crawl
glmc config show               # Display current configuration

CLI Output Schema

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "CLIOutput",
  "type": "object",
  "properties": {
    "execution_id": {"type": "string", "format": "uuid"},
    "timestamp": {"type": "string", "format": "date-time"},
    "query": {"type": "string"},
    "status": {
      "type": "string",
      "enum": ["success", "partial_success", "failure"]
    },
    "result": {
      "type": "object",
      "description": "Analysis result following AdvisoryOutput schema"
    },
    "metrics": {
      "type": "object",
      "properties": {
        "duration_ms": {"type": "number"},
        "tokens_used": {"type": "integer"},
        "gates_passed": {"type": "integer"},
        "degradation_level": {"type": "integer"}
      }
    },
    "quality_gates": {
      "type": "array",
      "items": {"$ref": "#/definitions/QualityGate"}
    }
  }
}

Knowledge Base Integration

Knowledge Entry Schema

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "KnowledgeEntry",
  "type": "object",
  "properties": {
    "id": {"type": "string"},
    "title": {"type": "string"},
    "authors": {"type": "array", "items": {"type": "string"}},
    "year": {"type": "integer"},
    "doi": {"type": "string", "format": "uri"},
    "tier": {
      "type": "string",
      "enum": ["Tier 1", "Tier 2", "Tier 3", "Tier 4"]
    },
    "relevance_score": {
      "type": "number",
      "minimum": 0,
      "maximum": 10
    },
    "keywords": {
      "type": "array",
      "items": {"type": "string"}
    },
    "summary": {"type": "string"},
    "key_findings": {
      "type": "array",
      "items": {"type": "string"}
    },
    "access_date": {"type": "string", "format": "date"}
  },
  "required": ["id", "title", "tier", "relevance_score"]
}

Extension Points

Custom Sub-Skills

To register a new sub-skill:

1. Create file in skills/sub-{name}.md
2. Add frontmatter with name and description
3. Implement required sections: Role, Workflow, Tools, Output, Quality Gates
4. Register in skills/main.md sub-skill catalog
5. Add input/output schema to this SKILL.md

Custom Quality Gates

To add a new quality gate:

1. Define gate ID (U7, G5, etc.)
2. Add check logic to tools/pipeline.py
3. Add auto-fix procedure if applicable
4. Update gate count in documentation
5. Add test case coverage

Performance Contracts

Latency Budget

| Component | Target | Max | |-----------|--------|-----| | Language Detection | 50ms | 100ms | | Requirements Gathering | 500ms | 2000ms | | Evidence Collection | 5000ms | 15000ms | | Core Analysis | 3000ms | 10000ms | | Knowledge Query | 1000ms | 3000ms | | Advisory Synthesis | 2000ms | 5000ms | | Quality Gate Check | 500ms | 2000ms | | Total Pipeline | 12s | 35s |

Resource Limits

{
  "resource_limits": {
    "max_context_tokens": 200000,
    "max_output_tokens": 4000,
    "max_file_size_mb": 10,
    "

…

## Source & license

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

- **Author:** [dungnotnull](https://github.com/dungnotnull)
- **Source:** [dungnotnull/game-livestream-multilingual-captioning-agent-skill](https://github.com/dungnotnull/game-livestream-multilingual-captioning-agent-skill)
- **License:** MIT

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