track: creative-apps
track_display_name: "🎨 Creative Apps"
notes: "Community Vote (10%) is excluded from automated scoring. Remaining 90% is prorated to 100%."
# === Scoring Enhancement: Evidence-Anchored Evaluation ===
# Scorers MUST cite specific evidence from the submission for each criterion.
# Generic statements like 'Comprehensive README' are NOT acceptable.
# Each score justification must reference concrete features, code patterns, or demos.
scoring_policy:
evidence_required: true
min_evidence_per_criterion: 1
differentiation_rules:
- "Score 8+ requires citing specific technical implementation details from README or code"
- "Score 9+ requires evidence of production-readiness or innovation beyond the basics"
- "Score 5-6 is the DEFAULT when evidence is ambiguous — do NOT default to 7+"
- "If two submissions have similar features, their scores MUST still differ based on depth/quality"
red_flags:
- signal: "No README or empty README"
max_ux_score: 4
- signal: "API keys or secrets visible in screenshots or code"
max_safety_score: 3
- signal: "No working demo and no clear setup instructions"
max_ux_score: 5
- signal: "Repository is private or inaccessible"
max_accuracy_score: 5
- signal: "Description is copy-paste from template with no customization"
max_creativity_score: 4
bonus_signals:
- signal: "Working deployed URL or live demo"
bonus_criteria: "UX & Presentation"
min_score: 7
- signal: "Automated tests present in repository"
bonus_criteria: "Reliability & Safety"
min_score: 7
- signal: "Multi-agent or MCP server architecture clearly documented"
bonus_criteria: "Accuracy & Relevance"
min_score: 7
criteria:
- name: "Accuracy & Relevance"
weight: 0.222
description: "Does the submission accurately meet challenge requirements? Degree of GitHub Copilot utilization."
scoring_guide:
"1-3": "Fails to meet more than half the requirements. Copilot usage unclear."
"4-6": "Meets basic requirements but partially incomplete. Copilot usage is limited."
"7-9": "Meets all requirements with effective Copilot usage. MCP integration present."
"10": "Exceeds requirements. Innovative Copilot usage. Multiple MCP integrations."
evidence_signals:
positive:
- "MCP server implementation found in code"
- "Copilot usage documented with specific examples"
- "All challenge requirements addressed with working features"
- "Multiple tool/MCP integrations visible in architecture"
negative:
- "Missing required features from challenge spec"
- "No evidence of Copilot usage"
- "Repository contains only boilerplate/template code"
- name: "Reasoning & Multi-step Thinking"
weight: 0.222
description: "Agent reasoning capability. Implementation of step-by-step thinking."
scoring_guide:
"1-3": "Simple input/output only. No reasoning steps."
"4-6": "Basic conditional branching. Limited multi-step reasoning."
"7-9": "Clear reasoning chain. Multi-step processing with error handling."
"10": "Self-improvement loop. Advanced reasoning patterns (ReAct/CoT) implemented."
evidence_signals:
positive:
- "Multi-step workflow visible in code or architecture diagram"
- "Error recovery or retry logic implemented"
- "Chain-of-Thought or ReAct pattern documented"
- "Agent decides next action based on previous results"
negative:
- "Single API call with no chaining"
- "Hardcoded responses without dynamic reasoning"
- "No error handling between steps"
- name: "Creativity & Originality"
weight: 0.167
description: "Originality of the idea. Value proposition beyond existing tools."
scoring_guide:
"1-3": "Simple wrapper around existing tools. No originality."
"4-6": "Some unique approaches. New solutions to known problems."
"7-9": "Unique concept. Clear differentiating factors."
"10": "Innovative. Creates a new category."
evidence_signals:
positive:
- "Novel combination of technologies not seen in other submissions"
- "Unique problem domain or approach"
- "Custom algorithms or processing pipelines"
- "Clear differentiator articulated in description"
negative:
- "Nearly identical to a common tutorial or template"
- "Problem already well-solved by existing tools"
- "No unique selling point identifiable"
- name: "UX & Presentation"
weight: 0.167
description: "User experience quality. Demo and documentation completeness."
scoring_guide:
"1-3": "No demo. Insufficient README. Usage unclear."
"4-6": "Screenshots provided. Basic README."
"7-9": "Video demo. Detailed README. Clear setup instructions."
"10": "Production-quality UX. Polished demo. Comprehensive documentation."
evidence_signals:
positive:
- "Video walkthrough showing actual usage"
- "Step-by-step setup guide that a new user can follow"
- "Architecture diagram included"
- "Error states and edge cases shown in demo"
negative:
- "README is auto-generated or default template"
- "Screenshots show only a terminal with no context"
- "No installation or setup instructions"
- name: "Reliability & Safety"
weight: 0.222
description: "Error handling. Security considerations. Robustness."
scoring_guide:
"1-3": "No error handling. API keys hardcoded."
"4-6": "Basic error handling. .env used but incomplete."
"7-9": "Comprehensive error handling. Security best practices followed."
"10": "Production quality. Rate limiting, retries, input validation complete."
evidence_signals:
positive:
- "Try/catch blocks around external API calls"
- ".env.example file with documented variables"
- "Input validation visible in code"
- "Rate limiting or retry logic implemented"
- "Automated tests present"
negative:
- "API keys visible in code or screenshots"
- "No .gitignore or secrets management"
- "Crashes on invalid input per demo"
