feat: Add interview-scorecard-calibrator kit

kits/interview-scorecard-calibrator/.env.example

@@ -0,0 +1,7 @@
+# Flow Configurations
+CALIBRATE_SCORECARD_FLOW_ID=your-lamatic-flow-uuid-here
+
+# API Credentials
+LAMATIC_API_KEY=your-lamatic-client-api-key-here
+LAMATIC_PROJECT_ID=your-lamatic-project-id-here
+LAMATIC_API_URL=https://your-organization-project-dev.lamatic.dev

kits/interview-scorecard-calibrator/.gitignore

@@ -0,0 +1,5 @@
+.lamatic/
+node_modules/
+.next/
+.env
+.env.local

kits/interview-scorecard-calibrator/README.md

@@ -0,0 +1,110 @@
+# Interview Scorecard Calibrator
+
+Reconcile multi-interviewer panel feedback into a calibrated hiring scorecard.
+
+## Problem
+
+After a panel interview, hiring managers receive fragmented notes — different formats, uneven scoring rigor, and conflicting opinions. Reconciling them into a fair hire/no-hire decision is slow and inconsistent.
+
+## Solution
+
+Paste a role rubric and interviewer notes. The Lamatic flow returns:
+
+- Per-competency calibrated scores with evidence
+- Interviewer disagreement map
+- Hire / lean-hire / lean-no / no-hire recommendation + confidence
+- Follow-up questions for missing evidence
+- Internal decision-summary email draft
+- A markdown hiring-committee brief
+
+## Why this kit is unique
+
+Unlike resume screeners (`hiring`, `hiring-copilot-agent`, `candidate-screener`) or candidate self-reflection coaches (`interview-reflection-coach`), this kit focuses on **post-panel calibration** for hiring managers.
+
+## Providers & Prerequisites
+
+- Lamatic account + deployed flow (import or rebuild from this kit)
+- LLM provider configured in Lamatic Studio (Gemini / OpenAI / etc.)
+- Node.js 20.9+
+
+## Setup in Lamatic Studio
+
+1. Create a project in [studio.lamatic.ai](https://studio.lamatic.ai)
+2. Recreate the flow from `flows/calibrate-scorecard.ts` (or import if your Studio supports kit import)
+3. Attach prompts from `prompts/`, model configs from `model-configs/`, and the normalize script from `scripts/`
+4. Bind a text generation model on both LLM nodes
+5. Deploy the flow and copy the **Flow ID**
+
+## Run locally
+
+```bash
+cd kits/interview-scorecard-calibrator/apps
+cp .env.example .env.local
+# fill in LAMATIC_API_URL, LAMATIC_PROJECT_ID, LAMATIC_API_KEY, CALIBRATE_SCORECARD_FLOW_ID
+npm install
+npm run dev
+```
+
+Open [http://localhost:3000](http://localhost:3000).
+
+## Environment variables
+
+Copy either kit-root `.env.example` or `apps/.env.example` into `apps/.env.local`.
+
+| Variable | Description |
+|---|---|
+| `LAMATIC_API_URL` | Lamatic API endpoint |
+| `LAMATIC_PROJECT_ID` | Project ID |
+| `LAMATIC_API_KEY` | API key |
+| `CALIBRATE_SCORECARD_FLOW_ID` | Deployed flow ID for `calibrate-scorecard` |
+
+## Interviewer notes format
+
+Provide **at least two** interviewer entries, separated by:
+
+- a line containing only `---`, and/or
+- headings like `Interviewer 1:` / `Interviewer 2 (Bob, Eng Manager):`
+
+Requests with fewer than two interviewer blocks are rejected before the model runs.
+
+## Sample input
+
+**Job title:** Senior Backend Engineer  
+**Level:** L5  
+
+**Rubric:**
+```
+System Design (High): scalable services, tradeoffs, reliability
+Coding (High): correctness, clarity, edge cases
+Ownership (Medium): end-to-end delivery, communication
+```
+
+**Interviewer notes:**
+```
+Interviewer 1 (Alice, Staff Eng):
+Strong system design around caching and failover. Coding solid but a bit slow. Ownership examples were concrete. Score: Design 4, Coding 3, Ownership 4.
+
+---
+
+Interviewer 2 (Bob, Eng Manager):
+Design felt hand-wavy on consistency. Coding was clean. Concerned about stakeholder communication. Score: Design 2, Coding 4, Ownership 2.
+```
+
+## Deploy
+
+Use the Vercel deploy link in `lamatic.config.ts` (`links.deploy`) with `root-directory=kits/interview-scorecard-calibrator/apps`.
+
+## Project structure
+
+```
+kits/interview-scorecard-calibrator/
+├── lamatic.config.ts
+├── agent.md
+├── README.md
+├── constitutions/default.md
+├── flows/calibrate-scorecard.ts
+├── prompts/
+├── model-configs/
+├── scripts/
+└── apps/                  # Next.js UI
+```

kits/interview-scorecard-calibrator/agent.md

@@ -0,0 +1,39 @@
+# Interview Scorecard Calibrator
+
+## Overview
+This kit helps hiring managers reconcile fragmented multi-interviewer feedback into a calibrated competency scorecard, disagreement map, hire/no-hire recommendation, and an internal decision-summary email draft. It is a single Lamatic flow invoked by the included Next.js app.
+
+## Purpose
+After panel interviews, notes arrive in different formats with uneven scoring rigor. Manual reconciliation is slow and inconsistent. This agent turns a shared rubric plus interviewer notes into a structured decision aid so committees can focus on evidence and disagreements instead of formatting.
+
+## Flows
+
+### `calibrate-scorecard`
+
+- Trigger: on-demand API execution from the kit UI or any Lamatic SDK caller
+- Inputs:
+  - `job_title` (required)
+  - `level` (optional)
+  - `rubric` (required) — competency list with weights/expectations
+  - `interviewer_notes` (required) — notes from 2–6 interviewers
+- Pipeline:
+  1. **Normalize Inputs** (`codeNode`) — validates and packages context
+  2. **Calibrate Scorecard** (`InstructorLLMNode`) — structured JSON scorecard
+  3. **Compose Brief** (`LLMNode`) — markdown hiring-committee brief
+  4. **API Response** — returns `scorecard` + `brief`
+- Outputs:
+  - `scorecard` — competencies, disagreements, recommendation, confidence, follow-ups, email draft
+  - `brief` — readable markdown summary for the hiring committee
+
+## Guardrails
+- Decision aid only — humans make the final hiring call
+- Do not fabricate interview evidence
+- Avoid protected-class inferences
+- Keep candidate/interviewer content confidential
+- See `constitutions/default.md`
+
+## Environment
+- `LAMATIC_API_URL`
+- `LAMATIC_PROJECT_ID`
+- `LAMATIC_API_KEY`
+- `CALIBRATE_SCORECARD_FLOW_ID`

kits/interview-scorecard-calibrator/apps/.env.example

@@ -0,0 +1,7 @@
+# Flow Configurations
+CALIBRATE_SCORECARD_FLOW_ID=your-lamatic-flow-uuid-here
+
+# API Credentials
+LAMATIC_API_KEY=your-lamatic-client-api-key-here
+LAMATIC_PROJECT_ID=your-lamatic-project-id-here
+LAMATIC_API_URL=https://your-organization-project-dev.lamatic.dev

kits/interview-scorecard-calibrator/apps/.gitignore

@@ -0,0 +1,23 @@
+/node_modules
+/.pnp
+.pnp.*
+.yarn/*
+!.yarn/patches
+!.yarn/plugins
+!.yarn/releases
+!.yarn/versions
+/coverage
+/.next/
+/out/
+/build
+.DS_Store
+*.pem
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+.pnpm-debug.log*
+.env*
+!.env.example
+.vercel
+*.tsbuildinfo
+next-env.d.ts

kits/interview-scorecard-calibrator/apps/README.md

@@ -0,0 +1,11 @@
+# Interview Scorecard Calibrator App
+
+Next.js UI for the Interview Scorecard Calibrator kit.
+
+```bash
+cp .env.example .env.local
+npm install
+npm run dev
+```
+
+Required env vars are documented in the parent [README.md](../README.md).

kits/interview-scorecard-calibrator/apps/actions/orchestrate.ts

@@ -0,0 +1,128 @@
+"use server";
+
+import { getLamaticClient } from "@/lib/lamatic-client";
+import { parseScorecard, splitInterviewerNotes, type Scorecard } from "@/lib/scorecard";
+import kitConfig from "../../lamatic.config";
+
+export type CalibrateInput = {
+  jobTitle: string;
+  level: string;
+  rubric: string;
+  interviewerNotes: string;
+};
+
+export type CalibrateResult = {
+  success: boolean;
+  data?: {
+    scorecard: Scorecard;
+    brief: string;
+    raw: unknown;
+  };
+  error?: string;
+};
+
+/** Resolve the deployed calibrate-scorecard flow ID from kit-root lamatic.config. */
+function getCalibrateFlowId() {
+  const step = kitConfig.steps.find((s) => s.id === "calibrate-scorecard");
+  if (!step?.envKey) {
+    throw new Error(
+      'lamatic.config has no step "calibrate-scorecard" with an envKey',
+    );
+  }
+
+  const flowId = process.env[step.envKey];
+  if (!flowId) {
+    throw new Error(
+      `${step.envKey} is not set. Set ${step.envKey} in the application environment.`,
+    );
+  }
+  return flowId;
+}
+
+/** Parse a JSON string when possible; otherwise return the original value. */
+function parseMaybeJson(value: unknown) {
+  if (typeof value !== "string") return value;
+  try {
+    return JSON.parse(value);
+  } catch {
+    return value;
+  }
+}
+
+/**
+ * Invokes the Lamatic calibrate-scorecard flow and validates the structured response.
+ */
+export async function calibrateScorecard(
+  input: CalibrateInput,
+): Promise<CalibrateResult> {
+  try {
+    const jobTitle = input.jobTitle.trim();
+    const rubric = input.rubric.trim();
+    const interviewerNotes = input.interviewerNotes.trim();
+    const level = input.level.trim();
+
+    if (!jobTitle) throw new Error("Job title is required");
+    if (!rubric) throw new Error("Competency rubric is required");
+    if (!interviewerNotes) throw new Error("Interviewer notes are required");
+
+    const interviewerBlocks = splitInterviewerNotes(interviewerNotes);
+    if (interviewerBlocks.length < 2) {
+      throw new Error(
+        "Provide notes from at least two interviewers, separated by '---' or 'Interviewer N:' headings",
+      );
+    }
+
+    const flowId = getCalibrateFlowId();
+    const lamaticClient = getLamaticClient();
+    const resData = await lamaticClient.executeFlow(flowId, {
+      job_title: jobTitle,
+      level: level || "Unspecified",
+      rubric,
+      interviewer_notes: interviewerNotes,
+    });
+
+    const result =
+      (resData as { result?: Record<string, unknown> })?.result || resData;
+    const scorecardRaw = parseMaybeJson(
+      (result as Record<string, unknown>)?.scorecard ??
+        (result as Record<string, unknown>)?.answer ??
+        result,
+    );
+    const scorecard = parseScorecard(scorecardRaw);
+    if (!scorecard) {
+      throw new Error(
+        "Flow returned an invalid scorecard payload. Check the deployed flow output schema.",
+      );
+    }
+
+    const brief =
+      typeof (result as Record<string, unknown>)?.brief === "string"
+        ? ((result as Record<string, unknown>).brief as string)
+        : "";
+
+    return {
+      success: true,
+      data: {
+        scorecard,
+        brief,
+        raw: resData,
+      },
+    };
+  } catch (error) {
+    let errorMessage = "Unknown error occurred";
+    if (error instanceof Error) {
+      errorMessage = error.message;
+      if (error.message.includes("fetch failed")) {
+        errorMessage =
+          "Network error: Unable to connect to Lamatic. Check your internet connection and API URL.";
+      } else if (error.message.toLowerCase().includes("api key")) {
+        errorMessage = "Authentication error: Please check your Lamatic API key.";
+      }
+    }
+
+    return {
+      success: false,
+      error: errorMessage,
+    };
+  }
+}

kits/interview-scorecard-calibrator/apps/app/globals.css

@@ -0,0 +1,55 @@
+@import "tailwindcss";
+@plugin "@tailwindcss/typography";
+
+:root {
+  --background: #f4f7fb;
+  --foreground: #0f172a;
+  --panel: #ffffff;
+  --accent: #0f766e;
+  --accent-soft: #ccfbf1;
+  --border: #dbe3ee;
+  --muted: #64748b;
+  --input: #e2e8f0;
+  --ring: #0f766e;
+  --primary: #0f766e;
+  --primary-foreground: #ffffff;
+  --secondary: #f1f5f9;
+  --secondary-foreground: #334155;
+  --destructive: #e11d48;
+  --rec-hire-bg: #d1fae5;
+  --rec-hire-fg: #065f46;
+  --rec-hire-border: #a7f3d0;
+  --rec-lean-hire-bg: #ccfbf1;
+  --rec-lean-hire-fg: #115e59;
+  --rec-lean-hire-border: #99f6e4;
+  --rec-lean-no-bg: #fef3c7;
+  --rec-lean-no-fg: #78350f;
+  --rec-lean-no-border: #fde68a;
+  --rec-no-hire-bg: #ffe4e6;
+  --rec-no-hire-fg: #9f1239;
+  --rec-no-hire-border: #fecdd3;
+  --rec-fallback-bg: #f1f5f9;
+  --rec-fallback-fg: #334155;
+  --rec-fallback-border: #e2e8f0;
+}
+
+@theme inline {
+  --color-background: var(--background);
+  --color-foreground: var(--foreground);
+  --font-sans: var(--font-source-sans);
+  --font-display: var(--font-fraunces);
+}
+
+body {
+  background:
+    radial-gradient(circle at top left, rgba(15, 118, 110, 0.12), transparent 32%),
+    radial-gradient(circle at top right, rgba(14, 165, 233, 0.1), transparent 28%),
+    linear-gradient(180deg, #eef5f8 0%, var(--background) 48%, #e8eef5 100%);
+  color: var(--foreground);
+  font-family: var(--font-sans), "Segoe UI", sans-serif;
+  min-height: 100vh;
+}
+
+.display-font {
+  font-family: var(--font-display), Georgia, serif;
+}