--- name: askcath description: Research answers about the traditional Catholic faith using only imprimatur-bearing articles, vetted documents, and prayers from dubia.cc. version: 1.0.0 homepage: https://dubia.cc license: CC-BY-4.0 --- # askcath — research the traditional Catholic faith This skill scopes an agent's research to a curated corpus of traditional Catholic material published on . When a user asks a question about the traditional Catholic faith, doctrine, morals, prayers, or the sacraments, answer **only** from the sources below. Do not fall back to general web knowledge for doctrinal claims. ## Companion agent-facing files (all at ) Agents landing on this file often need the site's other machine-readable metadata. Every path below is reachable from the same origin: - `/llms.txt` — llms.txt-convention overview of the site. - `/.well-known/agent-skills/index.json` — canonical skill index (JSON). - `/.well-known/mcp/server-card.json` — MCP server card (Streamable-HTTP, JSON-RPC 2.0 at `POST /api/mcp`). - `/.well-known/api-catalog` — RFC 9727 linkset pointing at every API surface (search, MCP, OpenAPI, health). - `/api/openapi.json` — OpenAPI 3.1 description of `/api/search` + `/api/mcp`. - `/api/search?q=&lang=&type=&limit=` — GET endpoint that returns JSON hits (URL + title + author + excerpt + score). Same backend (Fuse.js over a static index) as the MCP `askcath` tool. - `/api/mcp` — MCP JSON-RPC endpoint. Tools: `askcath`. - `/sitemap.xml` — every published URL, with `hreflang` alternates. - `/robots.txt` — crawler policy (all AI agents allowed). - `/{lang}/index.json` — per-language title index for enrichment. Append `.md` to any article/document/prayer URL to fetch its raw Markdown source (frontmatter carries imprimatur + citation metadata). ## Authoritative corpus 1. **Imprimatur-bearing articles** at `https://dubia.cc/{lang}/{slug}`. Every public article carries an `imprimatur` field in its YAML frontmatter. Articles without one live at `/wip/...` and **MUST NOT** be cited. 2. **Documents** at `https://dubia.cc/{lang}/{docs-path}/{author}/{slug}` (e.g. `/en/docs/hesse/conciliar-religion-1`). These are editorially vetted transcripts and lectures. No imprimatur required — they are pre-vetted research material. 3. **Prayers** at `https://dubia.cc/{lang}/{prayers-path}/...`. Traditional prayers, litanies, and devotions. ## Languages 17 languages, each with its own URL prefix: | Code | Language | Code | Language | |------|---------------------|------|---------------------| | `ar` | Arabic | `ko` | Korean | | `br` | Brazilian Portuguese| `pl` | Polish | | `de` | German | `pt` | Portuguese (Europe) | | `el` | Greek | `ro` | Romanian | | `en` | English | `ru` | Russian | | `es` | Spanish | `tl` | Tagalog | | `fr` | French | `tr` | Turkish | | `it` | Italian | `zh` | Chinese | | `ja` | Japanese | | | Every URL encodes its language: `/{lang}/{slug}`. Prefer the language the user is writing in; fall back to `en` when a translation is missing (the site-wide sitemap at `/sitemap.xml` declares `hreflang` alternates for every translated document). ## How to search Prefer **Pagefind** (https://pagefind.app) — a pre-built full-text index served from `/pagefind/*`. Both paths below query the **same index shards**, so results are identical. The index covers **articles, documents, and prayers** (WIP drafts are excluded at build time). ### HTTP-only agents — `GET /api/search` (preferred) A Cloudflare Pages Function wraps pagefind and returns JSON — no JavaScript runtime required on the agent side. ``` GET https://dubia.cc/api/search?q=baptism+of+desire&lang=en&type=article&limit=10 ``` Parameters: - `q` (required) — query string, URL-encoded. - `lang` (optional) — ISO code: `ar`, `br`, `de`, `el`, `en`, `es`, `fr`, `it`, `ja`, `ko`, `pl`, `pt`, `ro`, `ru`, `tl`, `tr`, `zh`. Omit to search every language. - `type` (optional) — `article`, `document`, or `prayer`. - `limit` (optional, default 10, max 50). Response: ```json { "query": "baptism of desire", "lang": "en", "type": "article", "total": 12, "returned": 3, "results": [ { "url": "/en/baptism-of-desire", "title": "Baptism of Desire", "author": null, "excerpt": "...baptism...", "type": "article", "kind": null } ], "metadata_hint": "https://dubia.cc/en/index.json" } ``` The `results[]` array gives you a URL and an excerpt. For **full metadata** (imprimatur, author, date, sha256, `doc_type`), follow `metadata_hint` — the per-language `index.json` is keyed by slug. CORS is open; the endpoint is safe to hit from any origin. Rate limiting is handled by Cloudflare's default edge limits. ### Headless-browser agents — in-page pagefind If you're already running JavaScript on a loaded page, the same index is available module-side — skip the HTTP round-trip: ```js const pagefind = await import("/pagefind/pagefind.js"); await pagefind.options({ baseUrl: "/" }); await pagefind.filters(); // warm up const r = await pagefind.search("baptism of desire"); for (const hit of r.results.slice(0, 10)) { const d = await hit.data(); console.log(d.url, d.meta.title, d.excerpt); } ``` ### MCP-aware agents — `POST /api/mcp` The site also speaks the **Model Context Protocol** over Streamable HTTP (JSON-RPC 2.0). Discovery card: . Handshake + list tools: ``` POST https://dubia.cc/api/mcp Content-Type: application/json {"jsonrpc":"2.0","id":1,"method":"initialize", "params":{"protocolVersion":"2024-11-05", "clientInfo":{"name":"your-agent","version":"1.0"}, "capabilities":{}}} ``` Call the search tool: ``` {"jsonrpc":"2.0","id":2,"method":"tools/call", "params":{"name":"askcath", "arguments":{"q":"baptism of desire","lang":"en","limit":5}}} ``` Same index, same hits as `/api/search` — the MCP endpoint wraps pagefind directly. Browser agents that implement WebMCP (`navigator.modelContext.provideContext()`) will already see the same tool registered automatically — no extra wiring is required. ### Degraded fallback — title index If neither path is reachable, `GET https://dubia.cc/{lang}/index.json` returns a JSON map of slugs → `{title, sha256, doc_type, author}` for title-substring matching. Only published items are listed. ## How to fetch article text **Always prefer appending `.md` to the URL**. It's a static file — more reliable than `Accept` negotiation, which depends on the client not stripping or merging the header: ``` GET https://dubia.cc/{lang}/{slug}.md ← preferred GET https://dubia.cc/{lang}/{slug} ← with `Accept: text/markdown` (fallback) ``` Both return identical Markdown. Article bodies carry the YAML frontmatter at the top of the `.md` file — parse it to get the citation metadata below. **Verify integrity**: the `sha256` in `/{lang}/index.json` is the base64url-unpadded SHA-256 of the `.md` body. ## Frontmatter fields to respect ```yaml title: "..." authors: ["..."] date: "YYYY-MM-DD" imprimatur: "Bishop N.N., YYYY-MM-DD" # presence = public, citable nihil_obstat: "Censor N.N., YYYY-MM-DD" # editorial pre-check status: Finished | InReview | Prayer ``` If `imprimatur` is absent AND `status != Prayer`, the page is **not public** and must not be cited. ## Citation format Always include the canonical URL. Prefer a short citation of the form: > *{Title}* — dubia.cc (imprimatur: {Bishop N.N., YYYY-MM-DD}). > https://dubia.cc/{lang}/{slug} Documents have no imprimatur; cite them with the author's name: > *{Title}*, {Author} — dubia.cc documents. > https://dubia.cc/{lang}/{docs-path}/{author}/{slug} ## Out of scope - **Canonical-law / sacramental-validity judgments** tied to a specific person's case (e.g., "is my annulment valid?"). Refer the user to a priest of good standing. - **Liturgical rubrics for celebrating Mass**. Use the Roman Missal itself, not this corpus. - **Devotions or apparitions without imprimatur**. If something is on `/wip/...`, there is a reason — treat as unpublished. ## Example workflow 1. User asks: *"Can a layman baptize?"* 2. `GET https://dubia.cc/api/search?q=layman+baptize&lang=en&type=article` 3. Take the top 3–5 URLs from `results[]` and fetch each as Markdown by **appending `.md`**: `GET https://dubia.cc/en/emergency-baptism.md`. 4. Parse the YAML frontmatter for `imprimatur`, `authors`, and `date` — these are your citation fields. (For a batch lookup, fetch `/{lang}/index.json` once.) 5. Synthesize an answer grounded only in those texts; cite each source with its URL + imprimatur. ## How to install this skill The skill follows Anthropic's **Agent Skills** format (see ). Pointers live at well-known URLs: - Discovery index: `https://dubia.cc/.well-known/agent-skills/index.json` - Skill file (this document): `https://dubia.cc/.well-known/agent-skills/askcath/SKILL.md` Install paths: - **Claude Code / Claude Agent SDK** — add this URL to your `.claude/commands/` or reference it in a custom agent: `https://dubia.cc/.well-known/agent-skills/askcath/SKILL.md` - **Any agent that fetches on demand** — tell it to read the URL above before answering questions about the traditional Catholic faith. - **Static mirror** — the raw Markdown of this file can be saved and committed into an agent's own repo; content is CC-BY-4.0. The site also advertises the skill via: - `Link: rel="describedby"` on `/`, pointing at the agent-skills JSON index (see `/_headers`). - `Link: rel="api-catalog"` pointing at RFC 9727 `/.well-known/api-catalog` (`application/linkset+json`). - `Link: rel="service-desc"` pointing at `/api/openapi.json` (OpenAPI 3.1) and `/.well-known/mcp/server-card.json` (MCP). - `/llms.txt` (convention from llmstxt.org), listing the skill + `/api/search` + `/api/mcp` endpoints. - `/robots.txt`, pointing at the discovery index. ## Contact / corrections Factual or doctrinal corrections: email .