# Developer Protocol **Server:** evals-mcp-server **Version:** 0.2.0 **Framework:** [@cyanheads/mcp-ts-core](https://www.npmjs.com/package/@cyanheads/mcp-ts-core) `^0.13.22` **Engines:** Bun ≥1.4.4, Node ≥24.0.0 **MCP SDK:** `@modelcontextprotocol/server` ^2.4.0 **Zod:** ^4.6.7 > **Read the framework docs first:** `node_modules/@cyanheads/mcp-ts-core/CLAUDE.md` contains the full API reference — builders, Context, error codes, exports, patterns. This file covers server-specific conventions only. --- ## This Server `evals-mcp-server` turns an agent into an **eval author**. It's an authoring-loop server (not a data-source wrapper): the agent produces a verifiable eval record through a **draft → review → surgical-revise → submit** loop, and the server acts as both scribe (normalize, persist, compile) and adversarial checker (run the record's own grader, reject what doesn't hold up). No external API — the server is the source of truth, backed by plain JSON files under `EVALS_DATA_DIR`. Full concept + as-built surface: [`docs/design.md`](docs/design.md), [`docs/idea.md`](docs/idea.md), [`docs/authoring-loop-pattern.md`](docs/authoring-loop-pattern.md). **Surface (9 tools, 1 record resource, 1 published skill):** | Primitive | Purpose | |:----------|:--------| | `evals_describe_schema` | Required/optional fields + grader options for a `task_type` (static). | | `evals_create_draft` | Create a draft; reflect it back with a review protocol + subagent prompt. | | `evals_get_record` | Read a draft or submitted record by id. | | `evals_revise_draft` | Surgical `set`/`append`/`unset` patch to a draft by dotted path. | | `evals_discard_draft` | Delete a draft. | | `evals_run_check` | Run a grader against candidates, PASS/REJECT, decoupled from any record. | | `evals_submit_draft` | Committability gate → freeze. | | `evals_list_records` | Filter records by status/domain/task_type/tag (summary projection). | | `evals_export_records` | Compile submitted records to JSONL/CSV/Inspect/lm-eval under `exports/`. | | `eval://record/{id}` | Resource mirror of `evals_get_record`. | | `skill://evals-guide/SKILL.md` | Published guide to independent verification, submission, and export. | **Services** (`src/services/`): `eval-record` (schema, draft builder, submit gate), `grader` (DSL execution + committability check, math.js for `numeric`), `record-store` (on-disk JSON CRUD, draft→submitted freeze, exports), `exporter` (the four output formats). **Load-bearing invariants:** records are a Zod `discriminatedUnion` on `task_type`; the grader DSL is a typed union serialized with the record. Deterministic submit requires gold and every positive to pass, at least one negative and all negatives rejected, and agreeing decorrelated verification. `llm_rubric` skips deterministic grading but requires a declared negative and verification. Draft readiness shares the record-local checks; capture resolution, confirmation, and deduplication run at submit. `captured_outputs` is server-owned and freshly resolved at submit. Storage is the `record-store` service, **not** `ctx.state`: validate persisted records on read and report invalid files explicitly. HTTP runs `stateful` only: `src/index.ts` declares `sessionMode: { require: 'stateful' }` because the `evals_submit_draft` confirmation is a multi-round `ctx.requestInput` a 2025-era HTTP client can answer only over a live session, so a stateless HTTP start is refused at boot — keep `Dockerfile` and `.env.example` on `stateful` to match. When changing tool surface or the schema, reconcile `docs/design.md`, `README.md`, and `server.json`. --- ## Core Rules - **Logic throws, framework catches.** Tool/resource handlers are pure — throw on failure, no `try/catch`. Plain `Error` is fine; the framework catches, classifies, and formats. Use error factories (`notFound()`, `validationError()`, etc.) when the error code matters. - **Use `ctx.log`** for request-scoped logging. No `console` calls. - **Use `ctx.state`** for tenant-scoped consent records. Eval records persist through the record-store service as human-inspectable JSON. - **Need input the caller didn't supply?** `return ctx.requestInput(...)` and read `ctx.inputs` when the handler is re-entered. Never `await` for user input mid-handler. - **Secrets in env vars only** — never hardcoded. - **Cut noise.** Add only what earns its place: no speculative generality, no guards for states the framework already prevents (Zod-validated params, classified errors), no abstraction until a third caller proves it, no option nothing sets. - **Close the loop on issues.** When implementing work tracked by a GitHub issue, comment on the issue with what landed and close it. Do both — a comment without a close leaves stale issues open; a close without a comment leaves no record of what shipped. The comment is for future readers — state the concrete changes, not the conversation that produced them. --- ## Patterns ### Tool ```ts import { tool, z } from '@cyanheads/mcp-ts-core'; import { JsonRpcErrorCode } from '@cyanheads/mcp-ts-core/errors'; import { type EvalRecord, RecordPayloadSchema } from '@/services/eval-record/schema.js'; import { getRecordStoreService } from '@/services/record-store/record-store-service.js'; export const getRecordTool = tool('evals_get_record', { title: 'evals-mcp-server: get record', description: 'Read a draft or submitted eval record by id. The verification subagent calls this to inspect a draft before re-deriving or looking up the gold.', annotations: { readOnlyHint: true, idempotentHint: true, openWorldHint: false }, input: z.object({ id: z.string().describe('The record id to read, e.g. "ev_7Qk2mNpXa".'), }), output: z.object({ record: RecordPayloadSchema.describe('The full eval record, including its grader and verification evidence.'), }), // Typed error contract — declares the domain failures this tool surfaces; the // record-store throws a notFound carrying this `not_found` reason, so the entry // is marked thrownBy: 'service' for the error-contract-unthrown lint. errors: [ { reason: 'not_found', code: JsonRpcErrorCode.NotFound, when: 'No record exists with the given id.', recovery: 'Use evals_list_records to browse existing record ids, then retry with a valid one.', thrownBy: 'service', }, ], async handler(input, ctx) { const record = await getRecordStoreService().require(input.id); ctx.log.debug('Read record', { id: record.id, status: record.status }); return { record }; }, // format() populates content[] — the markdown twin of structuredContent. // Different clients read different surfaces (Claude Code → structuredContent, // Claude Desktop → content[]); both must carry the same data. format: (result) => { const r = result.record as EvalRecord; return [{ type: 'text', text: `# ${r.id} (${r.status}) — ${r.task_type} · ${r.metadata.domain}` }]; }, }); ``` ### Resource ```ts import { resource, z } from '@cyanheads/mcp-ts-core'; import { RecordObjectSchema } from '@/services/eval-record/schema.js'; import { getRecordStoreService } from '@/services/record-store/record-store-service.js'; export const evalRecordResource = resource('eval://record/{id}', { name: 'eval-record', title: 'eval-record', description: 'A single draft or submitted eval record by id — the same payload evals_get_record returns.', mimeType: 'application/json', params: z.object({ id: z.string().describe('The record id, e.g. "ev_7Qk2mNpXa".') }), output: RecordObjectSchema, async handler(params, ctx) { // require() throws a notFound classified to the resource's typed error contract. const record = await getRecordStoreService().require(params.id); ctx.log.debug('Resolved record resource', { id: record.id, status: record.status }); return record; }, // list() enumerates resolvable URIs for resource-capable clients. list: async () => { const { summaries } = await getRecordStoreService().listSummaries({ limit: 100 }); return { resources: summaries.map((s) => ({ uri: `eval://record/${s.id}`, name: `${s.id} (${s.status})` })) }; }, }); ``` ### Server config ```ts // src/config/server-config.ts — lazy-parsed, separate from framework config import { z } from '@cyanheads/mcp-ts-core'; import { parseEnvConfig } from '@cyanheads/mcp-ts-core/config'; const ServerConfigSchema = z.object({ dataDir: z .string() .min(1) .default('./evals-data') // parseEnvConfig reads "", whitespace, and an unsubstituted ${…} as unset .describe('Root folder for record JSON. The store manages drafts/, submitted/, and exports/ under it.'), requireConfirmation: z .stringbool() .default(false) .describe('When true, evals_submit_draft requests human confirmation before finalizing.'), defaultLicense: z.string().optional().describe('Default metadata.license applied when a draft omits one.'), captureDir: z.string().optional().describe('Directory of tool-call capture files read at submit.'), }); let _config: z.infer | undefined; export function getServerConfig() { _config ??= parseEnvConfig(ServerConfigSchema, { dataDir: 'EVALS_DATA_DIR', requireConfirmation: 'EVALS_REQUIRE_CONFIRMATION', defaultLicense: 'EVALS_DEFAULT_LICENSE', captureDir: 'EVALS_CAPTURE_DIR', }); return _config; } ``` `parseEnvConfig` maps Zod schema paths → env var names so errors name the variable (`EVALS_DATA_DIR`) not the path (`dataDir`). Throws `ConfigurationError`, which the framework prints as a clean startup banner. For env booleans use `z.stringbool()`, never `z.coerce.boolean()` — `Boolean("false")` is `true`, so a coerced flag can't be disabled through the environment. `z.stringbool()` parses `true/false/1/0/yes/no/on/off` and rejects anything else, so `=false` actually disables. ### Server identity and instructions `createApp()` accepts optional identity fields forwarded to the SDK's `initialize` response and the server manifest (`/.well-known/mcp.json`): ```ts await createApp({ name: 'evals-mcp-server', title: 'evals-mcp-server', // must equal the unscoped package name — enforced by lint:packaging instructions: 'Author verifiable eval records through the draft → review → revise → submit loop. Call evals_describe_schema, then evals_create_draft; verify the gold in a fresh subagent, then evals_submit_draft.', // description is NOT set here — it derives from package.json (the canonical source); a copy is drift. // createApp() also accepts optional websiteUrl and icons forwarded to the initialize response. sessionMode: { require: 'stateful' }, // the submit confirmation needs a live session }); ``` `instructions` is optional server-level orientation, sent on every `initialize` as session-level context. Use it for deployment guidance (connection aliases, regional notes, scope hints) instead of repeating the same context across tool descriptions. Client adoption is uneven, but there's no downside when set. --- ## Context Handlers receive a unified `ctx` object. Key properties: | Property | Description | |:---------|:------------| | `ctx.log` | Request-scoped logger — `.debug()`, `.info()`, `.notice()`, `.warning()`, `.error()`. Auto-correlates requestId, traceId, tenantId. Dual-sink: Pino **and** `notifications/message` to the client, so treat it as client-visible. | | `ctx.state` | Tenant-scoped KV — `.get(key)`, `.set(key, value, { ttl? })`, `.delete(key)`, `.getMany(keys)`, `.list(prefix, { cursor, limit })`. Accepts any JSON-serializable value; reads return its JSON form (a `Date` comes back as an ISO string). | | `ctx.requestInput` | Suspend and ask the caller for more input — `return ctx.requestInput({ inputRequests: { key: inputRequired.elicit({ message, requestedSchema }) } })`. Never returns; the handler is re-entered with the answers. Always present. | | `ctx.inputs` | The request's responses — `.accepted(key, schema)`, `.view(key)`, `.state()`, `.dropped` — limited to what the client declared (`elicitation` and its form/url modes, `sampling`, `roots`). Client-supplied: a consent gate trusts only a `ctx.state` record it stored when it asked, bound to the operation, caller, and target (see the `api-context` skill). | | `ctx.clientCapabilities` | What the client declared for this request, `undefined` when no view exists. Decides whether to ask for optional context (e.g. roots); never a reason to skip a consent prompt. | | `ctx.enrich` | Success-path agent context (empty-result notices, query echo, pagination totals) — `ctx.enrich(...)` or `.notice()` / `.total()` / `.echo()` / `.truncated()`. Reaches `structuredContent` and `content[]`; lands only when the definition declares an `enrichment` block (no-op otherwise). | | `ctx.content` | Non-text content blocks — `.image(data, mimeType)`, `.audio(data, mimeType)`, or `ctx.content(block)` for a raw block. Prepended to `content[]` after `format()`; never enters `structuredContent`. | | `ctx.signal` | `AbortSignal` for cancellation. | | `ctx.requestId` | Request ID — the one every log record of the call carries and its error envelope returns as `data.requestId`. | | `ctx.tenantId` | Tenant ID from JWT; `'default'` for stdio or HTTP with auth off. | --- ## Errors Handlers throw — the framework catches, classifies, and formats. **Recommended: typed error contract.** Declare `errors: [{ reason, code, when, recovery, retryable?, severity?, thrownBy? }]` on `tool()` / `resource()` to receive `ctx.fail(reason, …)` typed against the reason union. TypeScript catches typos at compile time, `data.reason` is auto-populated for observability, linter enforces conformance against the handler body. `recovery` is required (≥ 5 words, lint-validated) — the single source of truth for the agent's next move. The framework puts it on the wire whenever a failure carrying that `reason` arrives without a hint — a bare `ctx.fail('reason')` or a service throw with `data: { reason }` — as `data.recovery.hint`, mirrored into `content[]` text unless the message already contains it verbatim; override with an explicit `{ recovery: { hint: '...' } }` when dynamic runtime context matters. Every error envelope also carries `data.requestId`, the id the server's log records for that call carry, and `content[]` closes with `(reason … · request )`. Mark an entry the service layer throws with `thrownBy: 'service'` so `error-contract-unthrown` skips it — lint-only metadata, nothing at runtime reads it. Baseline codes (`InternalError`, `ServiceUnavailable`, `Timeout`, `ValidationError`, `SerializationError`, `RequestCancelled`) bubble freely and don't need declaring. ```ts import { JsonRpcErrorCode } from '@cyanheads/mcp-ts-core/errors'; errors: [ { reason: 'no_match', code: JsonRpcErrorCode.NotFound, when: 'No item matched the query', recovery: 'Broaden the query or check the spelling and try again.' }, ], async handler(input, ctx) { const item = await db.find(input.id); if (!item) throw ctx.fail('no_match', `No item ${input.id}`); return item; } ``` **Declare contracts inline on each tool.** The contract is part of the tool's public surface — one file should give the full picture. Don't extract a shared `errors[]` constant; per-tool repetition is the intended cost of locality. **Fallback (no contract entry fits):** throw via factories or plain `Error`. ```ts // Error factories — explicit code import { notFound, serviceUnavailable } from '@cyanheads/mcp-ts-core/errors'; throw notFound('Item not found', { itemId }); throw serviceUnavailable('API unavailable', { url }, { cause: err }); // Plain Error — framework auto-classifies from message patterns throw new Error('Item not found'); // → NotFound throw new Error('Invalid query format'); // → ValidationError // McpError — when no factory exists for the code import { McpError, JsonRpcErrorCode } from '@cyanheads/mcp-ts-core/errors'; throw new McpError(JsonRpcErrorCode.InitializationFailed, 'Connection failed', { pool: 'primary' }); ``` See framework CLAUDE.md and the `api-errors` skill for the full auto-classification table, all available factories, and the contract reference. --- ## Structure ```text src/ index.ts # createApp() entry point config/ server-config.ts # Server-specific env vars (Zod schema) services/ [domain]/ [domain]-service.ts # Domain service (init/accessor pattern) types.ts # Domain types mcp-server/ tools/definitions/ [tool-name].tool.ts # Tool definitions resources/definitions/ [resource-name].resource.ts # Resource definitions prompts/definitions/ [prompt-name].prompt.ts # Prompt definitions ``` --- ## Naming | What | Convention | Example | |:-----|:-----------|:--------| | Files | kebab-case with suffix | `search-docs.tool.ts` | | Tool/resource/prompt names | snake_case | `search_docs` | | Directories | kebab-case | `src/services/doc-search/` | | Descriptions | Single string or template literal, no `+` concatenation | `'Search items by query and filter.'` | --- ## Skills Development skills live in `framework-skills/`. Read the matching skill directly or run `bun run list-skills` for the registry. Published skills live in root `skills/` and reach plugin installs, the npm package, Docker, and the MCPB bundle. To serve them over MCP, pass `skills: 'skills'` to `createApp()`. The `add-skill` skill covers authoring and delivery, and `lint:packaging` validates the tree. **Agent skill directory:** Copy skills into the directory your agent discovers (Claude Code: `.claude/skills/`, others: equivalent). Skills then load as context without referencing `framework-skills/` paths. After framework updates, run the `maintenance` skill — Phase B re-syncs the agent directory. Available skills: | Skill | Purpose | |:------|:--------| | `setup` | Post-init project orientation | | `design-mcp-server` | Design tool surface, resources, and services for a new server | | `add-tool` | Scaffold a new tool definition | | `add-app-tool` | Scaffold an MCP App tool + paired UI resource | | `add-resource` | Scaffold a new resource definition | | `add-prompt` | Scaffold a new prompt definition | | `add-skill` | Author and validate a published skill for agents that use this server | | `add-service` | Scaffold a new service integration | | `add-test` | Scaffold test file for a tool, resource, or service | | `field-test` | Exercise tools/resources/prompts with real inputs, verify behavior, report issues | | `tool-defs-analysis` | Read-only audit of MCP definition language across the surface — voice, leaks, defaults, recovery hints, output descriptions | | `security-pass` | Audit server for MCP-flavored security gaps: output injection, scope blast radius, input sinks, tenant isolation | | `code-simplifier` | Post-session cleanup against `git diff` — modernize syntax, consolidate duplication, align with the codebase | | `polish-docs-meta` | Finalize docs, README, metadata, and agent protocol for shipping | | `git-wrapup` | Land working-tree changes as a commit stack — version bump, changelog, verify, commit by concern, release commit on top. No tag, no push to main; opens the release PR when the project declares release PR mode | | `release-pr-review` | Review pass on an open release PR — simplifier + correctness review, fixes as ordinary commits on top of the stack, PR body kept in sync. Release PR mode only | | `release-and-publish` | Fast-forward merge (release PR mode) + tag + push + npm + MCP Registry + GH Release + Docker. Picks up from `git-wrapup` | | `maintenance` | Investigate changelogs, adopt upstream changes, sync skills to agent dirs | | `orchestrations` | Chain task skills into a gated multi-phase pipeline — build-out, QA-fix, update-ship — when you can spawn sub-agents | | `report-issue-framework` | File a bug or feature request against `@cyanheads/mcp-ts-core` via `gh` CLI | | `report-issue-local` | File a bug or feature request against this server's own repo via `gh` CLI | | `techniques` | Catalog of response/data-shaping techniques — overflow handling, payload shaping, retrieval patterns | | `api-auth` | Auth modes, scopes, JWT/OAuth | | `api-canvas` | DataCanvas: register tabular data, run SQL, export, plus the `spillover()` helper for big result sets — Tier 3 opt-in | | `api-config` | AppConfig, parseConfig, env vars | | `api-context` | Context interface, RequestContext, logger, state, multi-round-trip input | | `api-errors` | McpError, JsonRpcErrorCode, error patterns | | `api-linter` | Definition linter rule catalog — invoked by `bun run lint:mcp` and `devcheck` | | `api-mirror` | MirrorService: persistent self-refreshing local mirror (embedded SQLite + FTS5) of a bulk upstream dataset — Tier 3 opt-in | | `api-services` | LLM, Speech, Graph services | | `api-testing` | createMockContext, test patterns | | `api-utils` | Formatting, parsing, security, pagination, scheduling, telemetry helpers | | `api-telemetry` | OTel catalog: spans, metrics, completion logs, env config, cardinality rules | | `api-workers` | Cloudflare Workers runtime | **Chaining skills into pipelines.** When the user wants a multi-phase effort — build this server out, QA-and-fix the surface, update-and-ship — *and you can spawn sub-agents*, `framework-skills/orchestrations/SKILL.md` sequences the task skills above into a gated pipeline with verification at each step. Read it to drive the run. Optional: skip it if you can't orchestrate sub-agents, and ignore it entirely if you were *spawned* as one — you've already been scoped to a single phase. When you complete a skill's checklist, check the boxes and add a completion timestamp at the end (e.g., `Completed: 2026-03-11`). --- ## Commands **Runtime:** Scripts use Bun's native TypeScript execution — `bun run ` is the standard invocation. `npm run ` also works (npm delegates to bun). | Command | Purpose | |:--------|:--------| | `bun run build` | Compile TypeScript | | `bun run rebuild` | Clean + build | | `bun run clean` | Remove build artifacts | | `bun run devcheck` | Lint + format + typecheck + security + changelog sync | | `bun run audit:fix` | `bun audit fix` — upgrade vulnerable packages to the lowest safe version within existing ranges (`--dry-run` previews, `--latest` rewrites ranges). First response when `devcheck` flags a transitive advisory; then `bun update `, then `bun dedupe` | | `bun run audit:refresh` | Delete `bun.lock` and reinstall. Last resort after `audit:fix`, `bun update `, and `bun dedupe` — re-resolves every ranged dep (the framework pin included) and rewrites the lockfile as `lockfileVersion: 2` | | `bun run lint:mcp` | Run the MCP definition linter standalone (rule catalog: `api-linter` skill) | | `bun run lint:packaging` | Packaging surface checks — `server.json`/`manifest.json` env-var parity (run by devcheck) | | `bun run list-skills` | Print the skill registry | | `bun run tree` | Generate directory structure doc | | `bun run format` | Auto-fix formatting (safe fixes only) | | `bun run format:unsafe` | Also apply Biome's unsafe autofixes — review the diff; they can change behavior | | `bun run test` | Run tests (Vitest — use `bun run test`, not `bun test`) | | `bun run start:stdio` | Production mode (stdio) | | `bun run start:http` | Production mode (HTTP) | | `bun run changelog:build` | Regenerate `CHANGELOG.md` from `changelog/*.md` | | `bun run changelog:check` | Verify `CHANGELOG.md` is in sync (used by devcheck) | | `bun run bundle` | Build, pack, and clean a `.mcpb` for one-click Claude Desktop install | **CI is one file.** `.github/workflows/codeql.yml` (scaffolded) is the only GitHub Actions workflow: CodeQL is GitHub-owned end to end, and the file runs only while the repo's CodeQL *default setup* is turned off. Verification — `devcheck`, tests, the release gates — runs locally; don't add a workflow that re-runs it. --- ## Bundling `npm run bundle` produces a `.mcpb` extension bundle for Claude Desktop over stdio. Use root-anchored exclusions in `.mcpbignore` for source, tests, docs, caches, and local configuration. Root `skills/` ships in the bundle. After packing, `scripts/clean-mcpb.ts` prunes dev dependencies and strips dependency agent documentation, platform-specific bindings, node-gyp build output, and framework build-only files. The SDK's runtime skills modules stay in the bundle. **Adding an env var requires both files:** `server.json` (registry discovery, `environmentVariables[]`) and `manifest.json` (bundle install UX, `mcp_config.env` + `user_config`). `lint:packaging` (run by `devcheck`) verifies the env var names match, that every `user_config` option is wired into `mcp_config.env` as `"X": "${user_config.X}"` (the host substitutes nothing else — `"${X}"` reaches the server as that literal string), and that an optional string option carries `"default": ""`. **README install badges** (Claude Desktop `.mcpb`, Cursor, VS Code) and the `base64` / `encodeURIComponent` config-generation commands are ship-time concerns — run the `polish-docs-meta` skill, which carries the badge format, layout, and generation snippets in `framework-skills/polish-docs-meta/references/readme.md`. --- ## Changelog Directory-based, grouped by minor series via the `.x` semver-wildcard convention. Source of truth: `changelog/.x/.md` (e.g. `changelog/0.1.x/0.1.0.md`) — one file per release, shipped in the npm package. At release, author the per-version file with a concrete version and date, then run `npm run changelog:build` to regenerate the rollup. `changelog/template.md` is a **pristine format reference** — never edited or moved; read it for the frontmatter + section layout when scaffolding. `CHANGELOG.md` is a **navigation index** (header + link + summary per version), regenerated by `npm run changelog:build` — devcheck hard-fails on drift; never hand-edit it. Each per-version file opens with YAML frontmatter: ```markdown --- summary: "One-line headline, ≤350 chars" # required — powers the rollup index breaking: false # optional — true flags breaking changes security: false # optional — true ONLY for a source-code security fix, never a dependency CVE bump --- # 0.1.0 — YYYY-MM-DD ... ``` `breaking: true` renders a `· ⚠️ Breaking` badge — use it when consumers must update code on upgrade (signature changes, removed APIs, config renames). `security: true` renders a `· 🛡️ Security` badge and pairs with a `## Security` body section — set it only for a security fix in this server's *own source code*, never for a routine dependency or transitive CVE bump (record those under `## Dependencies`). When both are set, badges render `· ⚠️ Breaking · 🛡️ Security`. `agent-notes` is an optional free-form field for maintenance agents processing the release downstream. Content here won't appear in the rendered CHANGELOG — it's consumed by agents running the `maintenance` skill. Use it for adoption instructions that don't fit the human-facing sections: new files to create, fields to populate, one-time migration steps. Omit entirely when there's nothing to say. **Section order:** the Keep a Changelog sequence — Added, Changed, Deprecated, Removed, Fixed, Security — then `Dependencies` last. Include only sections with entries — don't ship empty headers. **Tag annotations** render as GitHub Release bodies via `--notes-from-tag`. They must be structured markdown — never a flat comma-separated string. Subject omits the version number (GitHub prepends it). See `changelog/template.md` for the full format reference. --- ## Publishing **Every release goes through a release PR, straight-through** — `git-wrapup`'s "Release PR mode", mode `straight-through`. One run: `git-wrapup` lands the commit stack on `release/`, pushes it, and opens the PR (title = the release commit subject, body = the release digest: theme line, `## Changes`, `## Gates`, changelog link last); `release-and-publish` then fast-forwards `main` locally with `git merge --ff-only`, creates the tag on `main`'s tip, pushes `main` and the tag, deletes the branch, and publishes. A caller's brief may run a given release as `gated` instead — a `release-pr-review` pass on the open PR before `release-and-publish`. **Never merge through the GitHub UI or `gh pr merge`**: squash and rebase-merge are disabled in the repo settings because both rewrite the stack (rebase-merge also strips the SSH signatures), and a merge commit breaks the linear history. --- ## Imports ```ts // Framework — z is re-exported, no separate zod import needed import { tool, z } from '@cyanheads/mcp-ts-core'; import { McpError, JsonRpcErrorCode } from '@cyanheads/mcp-ts-core/errors'; // Server's own code — via path alias import { getMyService } from '@/services/my-domain/my-service.js'; ``` --- ## Checklist - [ ] Zod schemas: all fields have `.describe()`, only JSON-Schema-serializable types (no `z.custom()`, `z.date()`, `z.transform()`, `z.bigint()`, `z.symbol()`, `z.void()`, `z.map()`, `z.set()`, `z.function()`, `z.nan()`) - [ ] Optional nested objects: handler guards for empty inner values from form-based clients (`if (input.obj?.field && ...)`, not just `if (input.obj)`). When regex/length constraints matter, use `z.union([z.literal(''), z.string().regex(...).describe(...)])` — literal variants are exempt from `describe-on-fields`. - [ ] JSDoc `@fileoverview` + `@module` on every file - [ ] `ctx.log` for logging, `ctx.state` for storage - [ ] Handlers throw on failure — error factories or plain `Error`, no try/catch - [ ] `format()` renders all data the LLM needs — different clients forward different surfaces (Claude Code → `structuredContent`, Claude Desktop → `content[]`); both must carry the same data - [ ] If wrapping external API: raw/domain/output schemas reviewed against real upstream sparsity/nullability before finalizing required vs optional fields - [ ] If wrapping external API: normalization and `format()` preserve uncertainty; do not fabricate facts from missing upstream data - [ ] If wrapping external API: tests include at least one sparse payload case with omitted upstream fields - [ ] Registered in `createApp()` arrays (directly or via barrel exports) - [ ] Tests use `createMockContext()` from `@cyanheads/mcp-ts-core/testing` - [ ] Root `plugin.json` and `mcp.json` follow Agent Plugins 1.0.0. Keep fixed environment values in root `mcp.json` and forward user variables through the Codex overlay's `env_vars`. The `mcpServers` pointer in `.codex-plugin/plugin.json` lets Codex merge that overlay. - [ ] `.codex-plugin/plugin.json` populated — `name`, `version`, `description`, `repository`, `license` from `package.json`; `interface.displayName` = the unscoped repo name (never the npm scope — `lint:packaging` enforces this); `interface.shortDescription` from `package.json` description - [ ] `.codex-plugin/mcp.json` updated — server name key is the unscoped repo name; every user-supplied variable (API key, contact email, instance URL) is listed in `env_vars` so Codex forwards it from the user's environment. Never write `"KEY": ""` into `env` — an empty value replaces the user's exported key and is read as unset - [ ] `.claude-plugin/plugin.json` populated — `name`, `version`, `description`, `author`, `repository`, `license`, `keywords` from `package.json`; inline `mcpServers` entry keyed by the unscoped repo name. Every user-supplied variable is declared under `userConfig` (`type`, `title`, `description`; `sensitive: true` for keys and tokens; `required: true` or `default: ""`) and referenced from `env` as `"KEY": "${user_config.