generated: '2026-08-25'
method: generated
source: openapi/ (9 documents, 28 operations) + conventions/, errors/, data-model/ in this repository
note: >-
  Packaged Agent Skills for the MBrace Therapeutics content API. Every operationId referenced in
  these skills was grepped out of the OpenAPI documents in openapi/ — none is invented. MBrace
  Therapeutics publishes no skills, no AGENTS.md and no agent documentation of its own; a search of
  mbracetrx.com on 2026-08-25 found none, so these are generated by API Evangelist from the harvested
  contract.
skills:
- name: mbrace-therapeutics-search-and-resolve
  file: mbrace-therapeutics-search-and-resolve.md
  description: >-
    Find any published object by keyword and resolve it to the full record without guessing integer
    identifiers. The entry-point skill — read this first.
  api: mbrace-therapeutics:mbrace-therapeutics-search-api
  operations: [searchContent, getNewsPost, getPage, getBoardMember, getExecutiveCommitteeMember, getFounder, getInvestor, getScientificBoardMember, getMediaItem]
- name: mbrace-therapeutics-track-news-archive
  file: mbrace-therapeutics-track-news-archive.md
  description: >-
    Read and poll the company news archive — press releases, conference presentations and
    publications — as structured JSON rather than scraped HTML.
  api: mbrace-therapeutics:mbrace-therapeutics-news-api
  operations: [listNews, getNewsPost, listCategories, getMediaItem]
- name: mbrace-therapeutics-harvest-company-graph
  file: mbrace-therapeutics-harvest-company-graph.md
  description: >-
    Harvest the board, executive committee, founders, named investors and scientific advisory board
    from the five custom post types MBrace registered. The marquee flow on this surface.
  api: mbrace-therapeutics:mbrace-therapeutics-people-api
  operations: [listBoardMembers, listExecutiveCommittee, listFounders, listInvestors, listScientificBoard, getMediaItem, listTypes]
shared_rules:
  authentication: none — all catalogued operations are anonymous GETs.
  pacing: One request per 10 seconds (robots.txt Crawl-delay), single concurrency, honour the 600-second cache.
  user_agent: Send a realistic browser User-Agent; an HTML body where JSON was expected is the Cloudflare edge challenge.
  error_handling: Branch on the `code` field, never the `message`. Do not retry a 401 — no self-service credential exists.
  field_selection: Always send _fields. The pages collection returns Elementor HTML bodies and is two orders of magnitude heavier without it.
  write_access: None. Every write route returns 401 anonymously and none is catalogued.