Speedrun Route Optimization Registry

Skill Registry Documentation

Overview

The speedrun-route-optimization microkernel v2.1.0 uses a dynamic skill registry with decorator-based registration, JSON Schema validation, and OpenAPI 3.0 specification generation.

Registered Skills

Main Harness

Name: speedrun-route-optimization Version: 2.1.0 Description: Main harness orchestrating 6-step workflow with quality gates Input Schema:

{
  "type": "object",
  "properties": {
    "user_input": {"type": "string"},
    "language": {"type": "string", "enum": ["en", "vi"]}
  },
  "required": ["user_input"]
}

Output Schema:

{
  "type": "object",
  "properties": {
    "verdict": {"type": "string"},
    "report": {"type": "string"}
  },
  "required": ["verdict"]
}

Sub-Skills

1. sub-gather-requirements

Version: 1.0.0 Tier: intake Quality Gate: G0 Description: Clarify game, category, target time, runner skill level, references, timing standard, language

2. sub-evidence-collector

Version: 1.0.0 Tier: evidence Description: Fetch current WR (speedrun.com), TAS reference (TASVideos), applicable glitches, cached academic refs

3. sub-core-analysis

Version: 1.0.0 Tier: analysis Quality Gates: G1, G2, G3 Description: DAG critical path + glitch replacement + frame-math feasibility + TAS-vs-RTA delta

4. sub-knowledge-updater

Version: 1.0.0 Tier: knowledge Quality Gate: G_knowledge Description: Surface 3-5 academic refs with Tier labels + gap flags + coverage rating

5. sub-advisor

Version: 1.0.0 Tier: synthesis Quality Gate: G_advisor Description: Synthesize into verdict + scenarios + risks + evidence chain + remediation + disclosure

Event System

Core Events

| Event | Payload | Handlers | |-------|---------|----------| | system.init | {version, config} | Core initialization | | skill.register | {skillname, version, schema} | Registry updates | | skill.pre-execution | {skillname, inputcontext} | Token tracking | | skill.post-execution | {skillname, output, durationms, tokens} | Metrics logging | | skill.error | {skillname, error, stacktrace} | Error recovery | | validation.success | {skillname, inputoroutput} | Metrics | | validation.failure | {skill_name, errors} | Error handling |

Registration API

from core.registry import register_skill

@register_skill(
    name="my-skill",
    version="1.0.0",
    description="My custom skill",
    input_schema={...},
    output_schema={...},
    subscribe_to=["skill.pre-execution"]
)
def my_skill(input_data):
    return {"result": "ok"}

OpenAPI Specification

OpenAPI 3.0 spec generated at: assets/ui/openapi.json Swagger UI available at: assets/ui/index.html

Generate documentation:

python scripts/generate_docs.py

Quality Gates

Universal Gates (U1-U6)

• U1: >=3 sources, >=1 academic
• U2: Disclosure before recommendation
• U3: Evidence tier per source
• U4: Language match
• U5: Template sections present
• U6: Claims traceable

Domain Gates (G1-G4)

• G1: Route decomposed (intended vs sequence-break)
• G2: Frame optimizations quantified
• G3: Alternatives with time/risk tradeoffs
• G4: Sources dated

Configuration

Configuration loaded from YAML files in config/:

• default.yaml - Base configuration
• production.yaml - Production overrides
• development.yaml - Development overrides

Load configuration:

from config import config
print(config.kernel_version)  # 2.1.0

Observability

Structured Logging

JSON format, component-based: speedrun.{component}

Token Tracking

Budget alerts at configurable threshold (default: 80%)

Metrics Export

Prometheus format on port 9090 (configurable)

Development

Setup

python scripts/dev_setup.py

Testing

pytest tests/

Documentation

python scripts/generate_docs.py

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/speedrun-route-optimization-agent-skill
• License: MIT

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