AI i agenci · Budowanie i kod
add-resource
/add-resource
Tworzy szkielet nowego zasobu MCP, który udostępnia dane pod adresem URI jako czytelny punkt dostępu.
Dla kogo
- Autor
- cyanheads ↗
- Repozytorium
- clinicaltrialsgov-mcp-server ↗
- Licencja
- Apache-2.0
- Ostatnia zmiana repozytorium
- 13.11.2026
Ocena i bezpieczeństwo
100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Przestarzałość
Brak sygnałów. Repozytorium zmienione 13.11.2026, bez odwołań do wycofanych modeli i API.
Licencja
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
Co robi i kiedy
Tworzy szkielet nowego zasobu MCP, który udostępnia dane pod adresem URI jako czytelny punkt dostępu.
Oryginalny opis (EN)
Scaffold a new MCP resource definition. Use when the user asks to add a resource, expose data via URI, or create a readable endpoint.
Jak działa
- Context
- Steps
- Template
- Checklist
Sekcje z nagłówków SKILL.md, w oryginale · 239 linii, 1 798 słów
Podgląd SKILL.md
---
name: add-resource
description: >
Scaffold a new MCP resource definition. Use when the user asks to add a resource, expose data via URI, or create a readable endpoint.
metadata:
author: cyanheads
version: "1.10"
audience: external
type: reference
---
## Context
Resources use the `resource()` builder from `@cyanheads/mcp-ts-core`. Each resource lives in `src/mcp-server/resources/definitions/` with a `.resource.ts` suffix. The standard registration pattern uses a `definitions/index.ts` barrel that collects all resources into an `allResourceDefinitions` array for `createApp()`. Fresh scaffolds start with direct imports in `src/index.ts` — the barrel is introduced as definitions grow. Match the pattern already used by the project you're editing.
**Tool coverage.** Not all MCP clients expose resources — many are tool-only (Claude Code, Cursor, most chat UIs). Before adding a resource, verify the same data is reachable via the tool surface — either through a dedicated tool, included in another tool's output, or bundled into a broader tool. A resource whose data has no tool path is invisible to a large share of agents.
## Steps
1. **Gather** the resource's URI template, purpose, and data shape from the user's request — ask only if genuinely absent
2. **Design the URI** — use `{paramName}` for path parameters (e.g., `myscheme://{itemId}/data`)
3. **Create the file** at `src/mcp-server/resources/definitions/{{resource-name}}.resource.ts`
4. **Register** the resource in the project's existing `createApp()` resource list (directly in `src/index.ts` for fresh scaffolds, or via a barrel if the repo already has one)
5. **Run `bun run devcheck`** to verify
6. **Smoke-test** with `bun run rebuild && bun run start:stdio < /dev/null` (or `start:http`) — the `Core services constructed` log record must list the new resource in its `resources` field (the message text shows only counts); if it doesn't, the resource never reached `createApp()`
## Template
```typescript
/**
* @fileoverview {{RESOURCE_DESCRIPTION}}
* @module mcp-server/resources/definitions/{{RESOURCE_NAME}}
*/
import { resource, z } from '@cyanheads/mcp-ts-core';
export const {{RESOURCE_EXPORT}} = resource('{{scheme}}://{{{paramName}}}/data', {
description: '{{RESOURCE_DESCRIPTION}}',
mimeType: 'application/json',
// size: 1024, // optional: content size in bytes, if known
params: z.object({
{{paramName}}: z.string().describe('{{PARAM_DESCRIPTION}}'),
}),
// auth: ['resource:{{resource_name}}:read'],
async handler(params, ctx) {
ctx.log.debug('Fetching resource', { {{paramName}}: params.{{paramName}} });
// Pure logic — throw on failure, no try/catch
return { /* resource data */ };
},
list: async (extra) => ({
resources: [
{
uri: '{{scheme}}://all',
name: '{{RESOURCE_LIST_NAME}}',
mimeType: 'application/json',
},
],
}),
});
```
### With pagination
For resources that return large result sets, use opaque cursor pagination in the `handler`. `resources/read` carries no cursor of its own, so the cursor must be a URI template variable — it arrives as a validated URI param. Make it a path segment: a `{?cursor}` query expansion is mandatory in the SDK's template matcher, so the bare URI (no `?cursor=`) stops matching. Serve the first page from an unpaged sibling resource that returns `nextCursor`, or disclose truncation in the body and point callers at a tool that pages. `paginateArray` requires a `RequestContext` for logging — create one from `requestContextService`:
```typescript
import { extractCursor, paginateArray, requestContextService } from '@cyanheads/mcp-ts-core/utils';
// URI template: '{{scheme}}://{{{paramName}}}/items/{cursor}'
params: z.object({
{{paramName}}: z.string().describe('{{PARAM_DESCRIPTION}}'),
cursor: z.string().optional().describe('Opaque pagination cursor'),
}),
async handler(params, ctx) {
const allItems = await fetchAllItems(params.{{paramName}});
const cursor = extractCursor({ cursor: params.cursor });
const reqCtx = requestContextService.createRequestContext({
operation: 'list-{{paramName}}',
parentContext: { requestId: ctx.requestId, traceId: ctx.traceId },
});
const page = paginateArray(allItems, cursor, 20, 100, reqCtx);
return {
items: page.items,
nextCursor: page.nextCursor,
};
},
```
### Registration
```typescript
// src/index.ts (fresh scaffold default)
import { createApp } from '@cyanheads/mcp-ts-core';
import { {{RESOURCE_EXPORT}} } from './mcp-server/resources/definitions/{{resource-name}}.resource.js';
await createApp({
tools: [/* existing tools */],
resources: [{{RESOURCE_EXPORT}}],
prompts: [/* existing prompts */],
});
```
If the repo already uses `src/mcp-server/resources/definitions/index.ts`, add the resource to that barrel the way it holds the existing ones — it must end up in the array passed to `createApp()`. A bare `export … from` line registers nothing on its own. The standard barrel shape:
```typescript
import { {{RESOURCE_EXPORT}} } from './{{resource-name}}.resource.js';
export const allResourceDefinitions = [/* existing resources */, {{RESOURCE_EXPORT}}];
```
### Optional: declarative `errors[]` contract
Resources can opt into the same typed error contract as tools — bound to a typed `ctx.fail(reason, …)` keyed by the declared reason union:
```typescript
import { JsonRpcErrorCode } from '@cyanheads/mcp-ts-core/errors';
export const articleResource = resource('article://{pmid}', {
description: 'Read an article by PMID.',
errors: [
{ reason: 'no_pmid_match', code: JsonRpcErrorCode.NotFound,
when: 'PMID not found in the index.',
recovery: 'Use pubmed_search_articles to discover valid PMIDs first.' },
{ reason: 'withdrawn', code: JsonRpcErrorCode.NotFound,
when: 'Article was withdrawn upstream.',
recovery: 'Check PubMed directly for retraction or withdrawal notices.' },
{ reason: 'upstream_throttled', code: JsonRpcErrorCode.RateLimited,
when: 'Upstream PubMed quota hit.', retryable: true,
recovery: 'Wait a few seconds and retry the request.' },
],
params: z.object({ pmid: z.string().describe('PubMed ID') }),
async handler(params, ctx) {
const article = await fetchOne(params.pmid);
if (!article) throw ctx.fail('no_pmid_match', `PMID ${params.pmid} not indexed`);
if (article.withdrawn) throw ctx.fail('withdrawn');
return article;
},
});
```
Without `errors[]`, the handler receives plain `Context` (no `fail` method) and throws via error factories (`notFound`, `serviceUnavailable`, …) directly. The contract is opt-in. See `framework-skills/api-errors/SKILL.md` for the full pattern, baseline codes, and conformance rules.
### URI template variable completion
Add a `complete` map to enable autocompletion for URI template variables. The SDK auto-installs `completion/complete` handling and advertises the `completions` capability when a registered resource template has completion callbacks — no other changes needed.
```typescript
import { resource, z } from '@cyanheads/mcp-ts-core';
const ITEM_IDS = ['item-001', 'item-002', 'item-abc'];
export const itemResource = resource('items://{itemId}', {
description: 'Retrieve an item by ID.',
params: z.object({ itemId: z.string().describe('Item identifier') }),
handler: (params) => ({ id: params.itemId }),
list: () => ({ resources: ITEM_IDS.map((id) => ({ uri: `items://${id}`, name: id })) }),
// Per-variable completion callbacks — keys must match URI template variable names.
complete: {
itemId: async (partial) => ITEM_IDS.filter((id) => id.startsWith(partial)),
},
});
```
Only applies to templated resources (URI templates with `{variable}` syntax). Static URIs don't support completion.
### Percent-encoded template variables
A handler receives each URI template variable exactly as the request URI carries it: reading `items://§ 141.61` hands it `itemId: '%C2%A7%20141.61'`, whether the client sent the URI encoded or raw. Opt in to decoding once, server-wide:
```typescript
await createApp({ resources: [itemResource], decodeResourceParams: true });
```
Each matched variable is then decoded once, before `params` validation and whether or not the definition declares `params`, inverting the SDK's encoding for its operator:
| Operator | Decoded with | `%2F` becomes |
|:---------|:-------------|:--------------|
| `{v}`, `{.v}`, `{/v}`, `{?v}`, `{&v}` | `decodeURIComponent` | `/` |
| `{+v}`, `{#v}` | `decodeURI` | `%2F` (reserved-character escapes stay encoded) |
Each element of an exploded variable (`{ids*}`) is decoded, and `+` stays `+`. A malformed escape (`%E0%A4%A`) refuses the read as a `params` rejection does: `-32007` with `data.issues` naming the variable. `ctx.uri` keeps the request URI as matched, and completion callbacks still receive the client's argument values. `createWorkerHandler` takes the same option.
Turning the switch on and deleting the handler's own `decodeURIComponent` are one change: a handler that keeps decoding decodes twice, so `100%2541` becomes `100A` and `100%25` makes its own decode throw.
A decoded value can hold `/`, `..`, `?`, and `#`: encode it again (`encodeURIComponent`) before placing it in an upstream URL, and constrain it with a `params` pattern before it names a file path, since `..` encodes to itself.
### Other `resource()` options
Beyond `description`, `params`, `handler`, and `list`, the builder also supports:
| Field | Purpose |
|:------|:--------|
| `name` | Short human-readable name for `resources/list`. Defaults to a slug derived from the URI template if omitted. |
| `output` | Optional Zod schema the handler's return is parsed against once, before `format()` (parity with `tool()`'s `output`): the handler returns the schema's input type (`z.input`), and `format()` receives the parsed value. |
| `format` | Optional formatter mapping the handler's return to the `ReadResourceResult.contents[]` shape. Default: string passthrough; objects serialized to JSON. Override when you need to attach permissions, custom encodings, or split into multiple content items. |
| `annotations` | Resource annotations (e.g., `audience`, `priority`) — see `ResourceAnnotations`. |
| `title` | Human-readable display title (defaults to `name`). On a template it names the template in `resources/templates/list`; the entries its `list()` returns never inherit it, so set `title` on an entry to give it one, or the client shows the entry's `name`. Those entries do inherit the template's `description`, `mimeType`, `annotations`, `size`, `examples`, and `_meta` unless they set their own. |
| `examples` | Array of `{ name, uri }` example entries surfaced in `resources/list` for discoverability. |
| `complete` | Per-variable completion callbacks for URI template variables. Keys match template variable names. Enables `completion/complete` and the `completions` capability. |
| `cacheHint` | `{ ttlMs?, cacheScope? }` — how long a client may cache this resource's `resources/read` result on protocol revision 2026-07-28. See below. |
### Cache hints (2026-07-28)
A resource that serves slow-changing data can declare how long its `resources/read` result stays fresh. `ttlMs` is the lifetime in milliseconds (a non-negative safe integer); `cacheScope` is `'private'` (only the requesting client may cache it) or `'public'` (shared caches may too).
```typescript
export const referenceTable = resource('reference://units', {
description: 'Unit conversion reference table.',
cacheHint: { ttlMs: 86_400_000, cacheScope: 'public' },
handler: () => UNIT_TABLE,
});
```
Resolution is per field, most specific first: the resource's own `cacheHint`, then the `resources/read` entry of `createApp({ cacheHints })`, then the SDK defaults (`ttlMs: 0`, `cacheScope: 'private'`). So a resource that names only a scope still inherits the server-wide lifetime.
Set a server-wide policy for the list operations alongside it — those results the SDK builds itself, so a resource cannot speak for them:
```typescript
await createApp({
cacheHints: {
'tools/list': { ttlMs: 3_600_000, cacheScope: 'public' },
'resources/read': { ttlMs: 60_000 },
},
resources: [referenceTable],
});
```
Cacheable operations are `tools/list`, `prompts/list`, `resources/list`, `resources/templates/list`, `resources/read`, and `server/discover`. Responses to 2025-era clients are unaffected — the hint only fills fields the 2026-07-28 revision defines.
## Checklist
- [ ] File created at `src/mcp-server/resources/definitions/{{resource-name}}.resource.ts`
- [ ] Resource name passed to `resource()` uses a valid URI template with `{paramName}` syntax
- [ ] All Zod `params` fields have `.describe()` annotations
- [ ] `output` schema added if the handler returns structured data that benefits from runtime validation
- [ ] JSDoc `@fileoverview` and `@module` header present
- [ ] `handler(params, ctx)` is pure — throws on failure, no try/catch
- [ ] If `errors[]` contract declared: every entry has a `recovery` field (≥5 words, lint-enforced)
- [ ] Data is reachable via the tool surface — confirm by checking `src/mcp-server/tools/definitions/` for a tool that exposes this data, or document why this resource is resources-only
- [ ] `list()` function provided if the resource is discoverable
- [ ] `cacheHint` set if the data is slow-changing and worth caching (`ttlMs` a non-negative safe integer)
- [ ] Pagination used for large result sets (`extractCursor`/`paginateArray`) — applies to both `handler` data and `list()` catalogs with many entries
- [ ] Registered in the project's existing `createApp()` resource list (directly or via barrel)
- [ ] `bun run devcheck` passes
- [ ] Smoke-tested with `bun run rebuild && bun run start:stdio < /dev/null` (or `start:http`); the `Core services constructed` record lists the new resource in its `resources` field
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · cyanheads/clinicaltrialsgov-mcp-server | 100 | Apache-2.0 |
| cyanheads/mcp-ts-core | 100 | Apache-2.0 |
| steamfield/stacks | 86 | Apache-2.0 |
| steamfield/stacks | 86 | Apache-2.0 |
| steamfield/stacks | 86 | Apache-2.0 |
Migawka bazy z 13.11.2026 · wersja paczki 4cf64b7df026 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 90mcp-resource-design
Udostępnia dane przez zasoby MCP ze stabilnymi URI, użytecznymi listami i rozmiarami mieszczącymi się w oknie kontekstu, gdy agent ma czytać dane. - 95add-tool
Tworzy szkielet nowej definicji narzędzia MCP dla serwera, gdy trzeba dodać nową funkcję. - 100add-app-tool
Tworzy szkielet pary: narzędzie MCP App i zasób UI, dla narzędzi z interaktywnym lub wizualnym interfejsem. - 91add-tool
Tworzy szkielet nowej definicji narzędzia MCP, gdy użytkownik chce dodać kolejną możliwość do serwera. - 91add-tool
Tworzy szkielet nowej definicji narzędzia MCP, gdy użytkownik chce dodać kolejną możliwość do serwera. - 100mcp-resources-uri-templates-subscriptions
Udostępnia dynamiczne pliki, wiersze baz danych i telemetrię przez zasoby MCP z szablonami URI oraz subskrypcjami na żywo.