Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Automate headless CMS operations in Prismic -- query documents, search content, retrieve custom types, and manage repository refs through the Composio Prismic integration.
.claude/skills/openteams-lab-prismic-automation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 161% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 189% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 460% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 155% | 0% |
| case-14 | ✓→✗ | ▼ Worse | 55% | 0% |
Manage your Prismic headless CMS directly from Claude Code. Query documents by type, full-text search content, inspect custom types, and work with repository refs for content versioning.
Toolkit docs: composio.dev/toolkits/prismic
https://rube.app/mcp
ref token. Always start by calling PRISMIC_REPOSITORY_API_GET_REFS or PRISMIC_REPOSITORY_API_GET_INFO to obtain the master ref.Retrieve comprehensive repository metadata including available refs (content versions), custom types, languages, tags, and bookmarks. This is typically your first API call.
Tools: PRISMIC_REPOSITORY_API_GET_INFO, PRISMIC_REPOSITORY_API_GET_REFS
No parameters required -- these endpoints return the full repository configuration. The refs field is critical since refs are required for all content queries.
Example prompt: "Get my Prismic repository info and the current master ref"
Fetch documents using Prismic's predicate query syntax with full pagination and filtering support.
Tool: PRISMIC_CONTENT_API_QUERY_DOCUMENTS
Key parameters:
ref (required) -- content release reference ID (typically the master ref)q -- predicate query, e.g., [[at(document.type, "page")]]page (min 1) and pageSize (1-100) -- paginationlang -- language code, e.g., en-us (default * for all)orderings -- sort order, e.g., [my.article.date desc]fetch -- comma-separated fields to fetch, reducing response sizefetchLinks -- resolve linked document fields inlineExample prompt: "Query all published blog posts in Prismic, sorted by date descending, in English"
Retrieve all documents of a specific custom type with automatic master ref resolution.
Tool: PRISMIC_GET_DOCUMENTS_BY_TYPE
Key parameters:
type (required) -- custom type API ID, e.g., blog_post, article, pagepage (default 1) and pageSize (1-100, default 20)lang -- language code filterorderings -- sort order, e.g., [my.article.date desc]after -- cursor-based pagination for deep pagination beyond page 50Example prompt: "Get all blog_post documents in Prismic, 20 per page"
Search across all text fields in documents for specified terms. Case-insensitive, matches on root words.
Tool: PRISMIC_CONTENT_API_GET_DOCUMENTS_WITH_FULLTEXT_SEARCH
Key parameters:
q (required) -- full-text predicate, e.g., [[fulltext(document, "machine learning")]]page, pageSize, lang, orderings -- same pagination/filtering as other queriesExample prompt: "Search all Prismic documents for 'machine learning'"
Retrieve a specific document by its unique identifier.
Tool: PRISMIC_GET_DOCUMENT_BY_ID
Key parameters:
document_id (required) -- unique document identifierref (required) -- content ref from repositorylang -- optional language filterExample prompt: "Fetch Prismic document Xx2KLhEAAJljVWaA"
Discover all custom types (content models) defined in the repository, including their structure definitions.
Tool: PRISMIC_TYPES_API_GET_TYPES
Key parameters:
limit -- max number of types to return per pagepage -- page number (1-indexed)sort -- sort order, e.g., nameExample prompt: "List all custom types in my Prismic repository"
ref (typically the master ref) from PRISMIC_REPOSITORY_API_GET_REFS or PRISMIC_REPOSITORY_API_GET_INFO before querying any documents. Queries without a ref will fail.[[at(document.type, "page")]]. For multiple predicates, combine them: [[at(document.type, "blog")][at(document.tags, ["featured"])]].after parameter with the last document ID from your previous result set.* (all languages). If you need documents in a specific locale, always pass lang explicitly (e.g., en-us, fr-fr).PRISMIC_CONTENT_API_GET_DOCUMENTS_WITH_INTEGRATION_FIELDS, you need an integrationFieldsRef in addition to the standard content ref.| Tool Slug | Description | |---|---| | PRISMIC_REPOSITORY_API_GET_INFO | Get repository metadata, refs, types, languages | | PRISMIC_REPOSITORY_API_GET_REFS | List all refs (master + releases) | | PRISMIC_TYPES_API_GET_TYPES | List all custom types / content models | | PRISMIC_CONTENT_API_QUERY_DOCUMENTS | Query documents with predicates and pagination | | PRISMIC_GET_DOCUMENTS_BY_TYPE | Fetch documents filtered by custom type | | PRISMIC_GET_DOCUMENT_BY_ID | Retrieve a single document by ID | | PRISMIC_CONTENT_API_GET_DOCUMENTS_WITH_FULLTEXT_SEARCH | Full-text search across all documents | | PRISMIC_CONTENT_API_GET_DOCUMENTS_WITH_PREDICATES | Filter documents with multiple predicate conditions | | PRISMIC_CONTENT_API_GET_DOCUMENTS_WITH_INTEGRATION_FIELDS | Fetch documents with integration fields data | | PRISMIC_GET_DOCUMENTS_ORDERED | Fetch documents sorted by specified fields |
Powered by Composio
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,983 | 4,191 | -68% | 1 | 1 | 0% | 2,603 | 1,975 | -24% | 0 | 0 | — |
case-02 | fail→fail | 13,723 | 3,608 | -74% | 1 | 1 | 0% | 1,720 | 1,919 | +12% | 0 | 0 | — |
case-03 | fail→fail | 7,277 | 7,493 | +3% | 1 | 1 | 0% | 1,303 | 2,608 | +100% | 0 | 0 | — |
case-04 | fail→fail | 6,504 | 8,631 | +33% | 1 | 1 | 0% | 1,159 | 3,188 | +175% | 0 | 0 | — |
case-14 | pass→fail | 7,148 | 4,675 | -35% | 1 | 1 | 0% | 1,332 | 2,060 | +55% | 0 | 0 | — |
case-05 | pass→pass | 7,074 | 4,423 | -37% | 1 | 1 | 0% | 1,324 | 2,449 | +85% | 0 | 0 | — |
case-06 | pass→pass | 9,986 | 5,969 | -40% | 1 | 1 | 0% | 1,811 | 2,754 | +52% | 0 | 0 | — |
case-07 | fail→fail | 8,173 | 7,698 | -6% | 1 | 1 | 0% | 1,571 | 2,176 | +39% | 0 | 0 | — |
case-08 | fail→pass | 8,063 | 10,865 | +35% | 1 | 1 | 0% | 1,359 | 3,542 | +161% | 0 | 0 | — |
case-09 | fail→fail | 9,693 | 4,058 | -58% | 1 | 1 | 0% | 1,776 | 2,347 | +32% | 0 | 0 | — |
case-10 | fail→fail | 6,378 | 8,518 | +34% | 1 | 1 | 0% | 1,167 | 2,504 | +115% | 0 | 0 | — |
case-11 | fail→fail | 9,486 | 5,358 | -44% | 1 | 1 | 0% | 1,812 | 1,930 | +7% | 0 | 0 | — |
case-12 | pass→fail | 8,709 | 5,523 | -37% | 1 | 1 | 0% | 1,524 | 2,093 | +37% | 0 | 0 | — |
case-13 | pass→pass | 9,432 | 5,751 | -39% | 1 | 1 | 0% | 1,690 | 2,684 | +59% | 0 | 0 | — |
case-15 | pass→pass | 8,464 | 5,074 | -40% | 1 | 1 | 0% | 1,313 | 2,491 | +90% | 0 | 0 | — |
case-16 | fail→fail | 7,571 | 5,005 | -34% | 1 | 1 | 0% | 1,388 | 1,924 | +39% | 0 | 0 | — |
case-17 | fail→pass | 4,584 | 2,575 | -44% | 1 | 1 | 0% | 712 | 2,057 | +189% | 0 | 0 | — |
case-18 | pass→fail | 4,733 | 5,682 | +20% | 1 | 1 | 0% | 860 | 2,206 | +157% | 0 | 0 | — |
case-19 | pass→pass | 6,507 | 4,529 | -30% | 1 | 1 | 0% | 1,104 | 2,371 | +115% | 0 | 0 | — |
case-20 | fail→pass | 3,773 | 8,621 | +128% | 1 | 1 | 0% | 548 | 3,068 | +460% | 0 | 0 | — |
case-21 | pass→pass | 4,564 | 4,093 | -10% | 1 | 1 | 0% | 685 | 2,326 | +240% | 0 | 0 | — |
case-22 | fail→pass | 5,913 | 5,107 | -14% | 1 | 1 | 0% | 1,012 | 2,583 | +155% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 14 counted toward the lift figure. The other 8 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +5 percentage points is the difference between those two pass rates over the 14 comparable cases. 4 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.