hackathon-event-parser（黑客马拉松事件解析器）

hackathon-event-parser

Goal

Parse a hackathon event URL to extract structured information — tracks, judging criteria, timeline, and sponsor tools — required to trigger the full devkit workflow pipeline autonomously.

This is the autonomous pipeline entry point. When a URL is provided, this skill replaces manual track description input for all downstream skills.

Trigger Conditions

Use this skill when:

• A hackathon event URL is available (Devpost, DoraHacks, Hackathon.com, MLH, or any event page)
• The team wants to run the devkit pipeline autonomously from a single input
• Track descriptions, judging criteria, or sponsor briefs must be extracted without manual copying
• Multiple tracks exist on an event page and the best-fit track must be identified
• This is always the first skill invoked in an autonomous pipeline; its output feeds hackathon-track-analyzer

Inputs

event_url: "<hackathon event URL>"           # required
preferred_track: "<track name or hint>"      # optional — if known, used to prioritize extraction
team_size: <number>                          # optional — used to filter feasibility signals
team_skills:                                 # optional — used to flag relevant sponsor tools
  - "<skill>"
extract_mode: "full | tracks_only | criteria_only"  # optional, default: full

Outputs

event_metadata:
  name: "<event name>"
  organizer: "<organizer name>"
  url: "<canonical event URL>"
  submission_platform: "<Devpost | DoraHacks | Hackathon.com | other>"
  start_datetime: "<ISO 8601 or 'unknown'>"
  end_datetime: "<ISO 8601 or 'unknown'>"
  duration_hours: <number or null>
  location: "<in-person | virtual | hybrid>"
  registration_deadline: "<ISO 8601 or 'unknown'>"
  submission_deadline: "<ISO 8601 or 'unknown'>"

tracks:
  - id: "<track-id>"
    name: "<track name>"
    description: "<raw track description>"
    sponsor: "<sponsor name or null>"
    prize: "<prize description or null>"
    required_tools:
      - "<tool or API name>"
    eligibility_constraints:
      - "<constraint>"
    feasibility_signal: "<high | medium | low>"  # based on team_skills if provided

judging_criteria:
  - track_id: "<track-id or 'global'>"
    axes:
      - axis: "<criterion name>"
        weight: "<high | medium | low | percentage if stated>"
        description: "<what judges evaluate>"
    rubric_source: "<verbatim | inferred | not_published>"

sponsor_tools:
  - sponsor: "<sponsor name>"
    tools:
      - name: "<tool or API>"
        use_case: "<what it enables>"
        bonus_prize: <true | false>
        docs_url: "<URL or null>"

timeline:
  - event: "<event name>"
    datetime: "<ISO 8601 or relative>"
    notes: "<optional context>"

recommended_track:
  track_id: "<id>"
  track_name: "<name>"
  rationale: "<why this track is recommended given team size/skills>"

extraction_confidence: "<high | medium | low>"
extraction_warnings:
  - "<any ambiguity, missing data, or access issues>"

next_skill: "hackathon-track-analyzer"

Rules

1. Extract all content directly from the event page without inference where possible.
2. If the judging rubric is not published, set rubric_source: "not_published" and infer typical criteria — mark inferred axes with [INFERRED].
3. If the URL is inaccessible (auth wall, 404, dynamic content), set extraction_confidence: "low" and populate fields from available fragments; list all gaps in extraction_warnings.
4. Never fabricate dates, prizes, or sponsor names. Use "unknown" or null for missing values.
5. If multiple tracks exist, extract all tracks and recommend one based on team signals if provided.
6. Normalize duration to hours. If only start/end dates are given, compute duration_hours.
7. Flag any required tool or API that conflicts with team_skills as a constraint.
8. next_skill must always be set to hackathon-track-analyzer to enforce pipeline continuity.

Output Format

event_metadata:
  name: "<string>"
  organizer: "<string>"
  url: "<string>"
  submission_platform: "<string>"
  start_datetime: "<ISO 8601 or 'unknown'>"
  end_datetime: "<ISO 8601 or 'unknown'>"
  duration_hours: <number or null>
  location: "<in-person | virtual | hybrid>"
  registration_deadline: "<ISO 8601 or 'unknown'>"
  submission_deadline: "<ISO 8601 or 'unknown'>"

tracks:
  - id: "<string>"
    name: "<string>"
    description: "<string>"
    sponsor: "<string or null>"
    prize: "<string or null>"
    required_tools:
      - "<string>"
    eligibility_constraints:
      - "<string>"
    feasibility_signal: "<high | medium | low>"

judging_criteria:
  - track_id: "<string>"
    axes:
      - axis: "<string>"
        weight: "<high | medium | low | percentage>"
        description: "<string>"
    rubric_source: "<verbatim | inferred | not_published>"

sponsor_tools:
  - sponsor: "<string>"
    tools:
      - name: "<string>"
        use_case: "<string>"
        bonus_prize: <boolean>
        docs_url: "<string or null>"

timeline:
  - event: "<string>"
    datetime: "<string>"
    notes: "<string>"

recommended_track:
  track_id: "<string>"
  track_name: "<string>"
  rationale: "<string>"

extraction_confidence: "<high | medium | low>"
extraction_warnings:
  - "<string>"

next_skill: "hackathon-track-analyzer"

Context Files

Knowledge Base

• knowledge/hackathon-tools.md
• knowledge/hackathon-judging-criteria.md
• knowledge/hackathon-winning-patterns.md
• knowledge/hackathon-submission-guidelines.md

Playbooks

• playbooks/hackathon-workflow.md
• playbooks/24h-hackathon-playbook.md
• playbooks/36h-hackathon-playbook.md
• playbooks/48h-hackathon-playbook.md

Input:

event_url: "https://globalaihealth.devpost.com"
team_size: 3
team_skills: ["Python", "FastAPI", "React", "OpenAI API"]
extract_mode: "full"

Output:

event_metadata:
  name: "Global AI Health Hackathon 2025"
  organizer: "HealthTech Alliance"
  url: "https://globalaihealth.devpost.com"
  submission_platform: "Devpost"
  start_datetime: "2025-09-12T09:00:00-05:00"
  end_datetime: "2025-09-14T17:00:00-05:00"
  duration_hours: 56
  location: "virtual"
  registration_deadline: "2025-09-11T23:59:00-05:00"
  submission_deadline: "2025-09-14T15:00:00-05:00"

tracks:
  - id: "track-mental-health"
    name: "AI for Mental Health"
    description: "Build AI-powered tools that improve mental health outcomes for young adults aged 18–30."
    sponsor: "OpenAI"
    prize: "$5,000 + OpenAI API credits"
    required_tools:
      - "OpenAI API"
    eligibility_constraints:
      - "Must use OpenAI API"
      - "Solution must target ages 18–30"
    feasibility_signal: "high"

  - id: "track-diagnostics"
    name: "AI Diagnostics"
    description: "Use AI to assist clinicians in early disease detection using medical imaging."
    sponsor: "Nvidia"
    prize: "$7,500 + Nvidia GPU credits"
    required_tools:
      - "Nvidia NIM"
      - "Medical imaging dataset"
    eligibility_constraints:
      - "Must use a publicly available medical dataset"
      - "Model must not make clinical claims"
    feasibility_signal: "low"

judging_criteria:
  - track_id: "track-mental-health"
    axes:
      - axis: "Innovation"
        weight: "30%"
        description: "Novel application of AI to mental health — beyond basic chatbots"
      - axis: "Impact"
        weight: "30%"
        description: "Potential to reach underserved users; measurable outcomes"
      - axis: "Technical Execution"
        weight: "20%"
        description: "Working demo; appropriate use of OpenAI API; code quality"
      - axis: "Presentation"
        weight: "20%"
        description: "Clarity of pitch; demo quality; storytelling"
    rubric_source: "verbatim"

  - track_id: "track-diagnostics"
    axes:
      - axis: "Clinical Validity"
        weight: "high"
        description: "[INFERRED] Accuracy and safety of diagnostic suggestions"
      - axis: "Technical Execution"
        weight: "high"
        description: "[INFERRED] Model performance; dataset usage; implementation quality"
    rubric_source: "inferred"

sponsor_tools:
  - sponsor: "OpenAI"
    tools:
      - name: "GPT-4o API"
        use_case: "Conversational AI, text generation, classification"
        bonus_prize: true
        docs_url: "https://platform.openai.com/docs"
  - sponsor: "Nvidia"
    tools:
      - name: "Nvidia NIM"
        use_case: "GPU-accelerated model inference for medical imaging models"
        bonus_prize: true
        docs_url: "https://developer.nvidia.com/nim"

timeline:
  - event: "Registration closes"
    datetime: "2025-09-11T23:59:00-05:00"
    notes: "Teams must register before hacking begins"
  - event: "Hackathon start"
    datetime: "2025-09-12T09:00:00-05:00"
    notes: "Opening keynote; tracks announced"
  - event: "Midpoint check-in"
    datetime: "2025-09-13T12:00:00-05:00"
    notes: "Optional mentor sessions available"
  - event: "Submissions close"
    datetime: "2025-09-14T15:00:00-05:00"
    notes: "Devpost submission must be complete; late submissions not accepted"
  - event: "Judging"
    datetime: "2025-09-14T16:00:00-05:00"
    notes: "3-minute pitch + 2-minute Q&A per team"

recommended_track:
  track_id: "track-mental-health"
  track_name: "AI for Mental Health"
  rationale: "Team has Python, FastAPI, React, and OpenAI API skills — all required tools are covered. Feasibility signal is high vs. low for diagnostics (which requires medical imaging expertise). Prize is competitive and judging rubric is fully published."

extraction_confidence: "high"
extraction_warnings:
  - "Judging rubric for track-diagnostics was not published; criteria are inferred from track description"
  - "Midpoint check-in schedule may change — verify on event Discord"

next_skill: "hackathon-track-analyzer"