research-to-json

name	research-to-json
description	Transform candidate research markdown files from research/ into deterministic per-candidate JSON files in data/. Use this skill whenever a user asks to convert, structure, normalize, or prepare candidate research for seeding/Payload ingestion.

Research Markdown -> Candidate JSON

Use this skill to convert research/*.md files into structured candidate JSON files for seeding.

Goal

Produce one JSON file per candidate in data/:

• research/ivan-cepeda.md -> data/ivan-cepeda.json
• research/abelardo-de-la-espriella.md -> data/abelardo-de-la-espriella.json

Do not generate one big multi-candidate JSON document.

Output contract

Write a JSON object that maps directly to the Candidates collection fields:

{
  "name": "string",
  "slug": "string",
  "party": "string",
  "currentOffice": "string",
  "lastUpdated": "YYYY-MM-DD",
  "socialLinks": [
    { "platform": "x|instagram|facebook|youtube", "url": "https://..." }
  ],
  "biography": "string",
  "publicTrajectoryItems": [
    {
      "role": "string",
      "organization": "string",
      "startYear": "string",
      "endYear": "string",
      "location": "string",
      "description": "string"
    }
  ],
  "privateTrajectoryItems": [
    {
      "role": "string",
      "organization": "string",
      "startYear": "string",
      "endYear": "string",
      "location": "string",
      "description": "string"
    }
  ],
  "proposals": "string",
  "proposalItems": [
    {
      "title": "string",
      "description": "string",
      "topic": "string",
      "sourceTitle": "string",
      "sourceUrl": "https://...",
      "sourceTier": "oficial|prensa|ong|redes"
    }
  ],
  "controversies": "string",
  "controversyItems": [
    {
      "title": "string",
      "description": "string",
      "status": "suspicion|under_investigation|indicted|cleared|convicted",
      "year": "string",
      "sourceTitle": "string",
      "sourceUrl": "https://...",
      "sourceTier": "oficial|prensa|ong|redes"
    }
  ],
  "alliances": "string",
  "allianceParties": [{ "name": "string" }],
  "endorsers": [{ "name": "string" }],
  "record": "string",
  "funding": "string",
  "sources": [
    {
      "section": "biography|proposals|controversies|alliances|record|funding",
      "title": "string",
      "publishedAt": "YYYY-MM-DD",
      "url": "https://...",
      "tier": "oficial|prensa|ong|redes"
    }
  ],
  "summaryTrajectory": "string",
  "summaryProposals": "string",
  "summaryControversies": "string",
  "summaryAlliances": "string",
  "summaryRecord": "string",
  "summaryFunding": "string"
}

Mapping rules

1. Read the research markdown headings and map by explicit CMS field names in backticks.
2. Parse code-block list items (1., 2., etc.) into arrays preserving order.
3. Keep wording faithful to the research file; do not invent or embellish.
4. Normalize keys to exact field names from the contract above.
5. For optional arrays with no entries, use [].
6. Keep dates in YYYY-MM-DD where present.
7. Preserve Spanish text exactly (except minimal cleanup for broken whitespace).
8. Exclude operational notes (e.g., "subir manualmente") from JSON values.

Quality checks before writing

• slug must match output filename.
• socialLinks[].platform must be one of: x, instagram, facebook, youtube.
• sourceTier and sources[].tier must be one of: oficial, prensa, ong, redes.
• controversyItems[].status must be one of: suspicion, under_investigation, indicted, cleared, convicted.
• sources[].section must be one of: biography, proposals, controversies, alliances, record, funding.

Behavior expectations

• If a field is truly missing in research markdown, leave it as an empty string or empty array (whichever matches type) and mention it in your final note.
• Never insert opinions, rankings, or sentiment language.
• Keep output deterministic so future runs over the same markdown produce the same JSON shape.