SKILL.md
$27
Run these commands to understand the project before making recommendations:
# Confirm NestJS project
grep -E '"@nestjs/core"' package.json 2>/dev/null
# Check NestJS version
node -e "console.log(require('./node_modules/@nestjs/core/package.json').version)" 2>/dev/null
# Check existing Sentry
grep -i sentry package.json 2>/dev/null
ls src/instrument.ts 2>/dev/null
grep -r "Sentry.init\|@sentry" src/main.ts src/instrument.ts 2>/dev/null
# Check for existing Sentry DI wrapper (common in enterprise NestJS)
grep -rE "SENTRY.*TOKEN|SentryProxy|SentryService" src/ libs/ 2>/dev/null
# Check for config-class-based init (vs env-var-based)
grep -rE "class SentryConfig|SentryConfig" src/ libs/ 2>/dev/null
# Check if SentryModule.forRoot() is already registered in a shared module
grep -rE "SentryModule\.forRoot|SentryProxyModule" src/ libs/ 2>/dev/null
# Detect HTTP adapter (default is Express)
grep -E "FastifyAdapter|@nestjs/platform-fastify" package.json src/main.ts 2>/dev/null
# Detect GraphQL
grep -E '"@nestjs/graphql"|"apollo-server"' package.json 2>/dev/null
# Detect microservices
grep '"@nestjs/microservices"' package.json 2>/dev/null
# Detect WebSockets
grep -E '"@nestjs/websockets"|"socket.io"' package.json 2>/dev/null
# Detect task queues / scheduled jobs
grep -E '"@nestjs/bull"|"@nestjs/bullmq"|"@nestjs/schedule"|"bullmq"|"bull"' package.json 2>/dev/null
# Detect databases
grep -E '"@prisma/client"|"typeorm"|"mongoose"|"pg"|"mysql2"' package.json 2>/dev/null
# Detect AI libraries
grep -E '"openai"|"@anthropic-ai"|"langchain"|"@langchain"|"@google/generative-ai"|"ai"' package.json 2>/dev/null
# Check for companion frontend
ls -d ../frontend ../web ../client ../ui 2>/dev/null
What to note:
- Is
@sentry/nestjsalready installed? If yes, check ifinstrument.tsexists andSentry.init()is called — may just need feature config.
- Sentry DI wrapper detected? → The project wraps Sentry behind a DI token (e.g.
SENTRY_PROXY_TOKEN) for testability. Use the injected proxy for all runtime Sentry calls (startSpan,captureException,withIsolationScope) instead of importing@sentry/nestjsdirectly in controllers, services, and processors. Onlyinstrument.tsshould import@sentry/nestjsdirectly.
- Config class detected? → The project uses a typed config class for
Sentry.init()options (e.g. loaded from YAML or@nestjs/config). Any new SDK options must be added to the config type — do not hardcode values that should be configurable per environment.
- **
SentryModule.forRoot()already registered?** → If it's in a shared module (e.g. a Sentry proxy module), do not add it again inAppModule— this causes duplicate interceptor registration.
- Express (default) or Fastify adapter? Express is fully supported; Fastify works but has known edge cases.
- GraphQL detected? →
SentryGlobalFilterhandles it natively.
- Microservices detected? → Recommend RPC exception filter.
- Task queues /
@nestjs/schedule? → Recommend crons.
- AI libraries? → Auto-instrumented, zero config.
- Prisma? → Requires manual
prismaIntegration().
- Companion frontend? → Triggers Phase 4 cross-link.
Phase 2: Recommend
Based on what you found, present a concrete proposal. Don't ask open-ended questions — lead with a recommendation:
Always recommended (core coverage):
- ✅ Error Monitoring — captures unhandled exceptions across HTTP, GraphQL, RPC, and WebSocket contexts
- ✅ Tracing — auto-instruments middleware, guards, pipes, interceptors, filters, and route handlers
Recommend when detected:
- ✅ Profiling — production apps where CPU performance matters (
@sentry/profiling-node)
- ✅ Logging — structured Sentry Logs + optional console capture
- ✅ Crons —
@nestjs/schedule, Bull, or BullMQ detected
- ✅ Metrics — business KPIs or SLO tracking
- ✅ AI Monitoring — OpenAI/Anthropic/LangChain/etc. detected (auto-instrumented, zero config)
Recommendation matrix:
Feature
Recommend when...
Reference
Error Monitoring
Always — non-negotiable baseline
${SKILL_ROOT}/references/error-monitoring.md
Tracing
Always — NestJS lifecycle is auto-instrumented
${SKILL_ROOT}/references/tracing.md
Profiling
Production + CPU-sensitive workloads
${SKILL_ROOT}/references/profiling.md
Logging
Always; enhanced for structured log aggregation
${SKILL_ROOT}/references/logging.md
Metrics
Custom business KPIs or SLO tracking
${SKILL_ROOT}/references/metrics.md
Crons
@nestjs/schedule, Bull, or BullMQ detected
${SKILL_ROOT}/references/crons.md
AI Monitoring
OpenAI/Anthropic/LangChain/etc. detected
${SKILL_ROOT}/references/ai-monitoring.md
Propose: "I recommend Error Monitoring + Tracing + Logging. Want Profiling, Crons, or AI Monitoring too?"
Phase 3: Guide
Install
# Core SDK (always required — includes @sentry/node)
npm install @sentry/nestjs
# With profiling support (optional)
npm install @sentry/nestjs @sentry/profiling-node
⚠️ **Do NOT install @sentry/node alongside @sentry/nestjs** — @sentry/nestjs re-exports everything from @sentry/node. Installing both causes duplicate registration.
Three-File Setup (Required)
NestJS requires a specific three-file initialization pattern because the Sentry SDK must patch Node.js modules (via OpenTelemetry) before NestJS loads them.
Before creating new files, check Phase 1 results:
- If
instrument.tsalready exists → modify it, don't create a new one.
- If a config class drives
Sentry.init()→ read options from the config instead of hardcoding env vars.
- If a Sentry DI wrapper exists → use it for runtime calls instead of importing
@sentry/nestjsdirectly in services/controllers.
#### Step 1: Create src/instrument.ts
import * as Sentry from "@sentry/nestjs";
// Optional: add profiling
// import { nodeProfilingIntegration } from "@sentry/profiling-node";
Sentry.init({
dsn: process.env.SENTRY_DSN,
environment: process.env.SENTRY_ENVIRONMENT ?? "production",
release: process.env.SENTRY_RELEASE,
sendDefaultPii: true,
// Tracing — lower to 0.1–0.2 in high-traffic production
tracesSampleRate: 1.0,
// Profiling (requires @sentry/profiling-node)
// integrations: [nodeProfilingIntegration()],
// profileSessionSampleRate: 1.0,
// profileLifecycle: "trace",
// Structured logs (SDK ≥ 9.41.0)
enableLogs: true,
});
**Config-driven Sentry.init():** If Phase 1 found a typed config class (e.g. SentryConfig), read options from it instead of using raw process.env. This is common in NestJS apps that use @nestjs/config or custom config loaders:
import * as Sentry from "@sentry/nestjs";
import { loadConfiguration } from "./config";
const config = loadConfiguration();
Sentry.init({
dsn: config.sentry.dsn,
environment: config.sentry.environment ?? "production",
release: config.sentry.release,
sendDefaultPii: config.sentry.sendDefaultPii ?? true,
tracesSampleRate: config.sentry.tracesSampleRate ?? 1.0,
profileSessionSampleRate: config.sentry.profilesSampleRate ?? 1.0,
profileLifecycle: "trace",
enableLogs: true,
});
When adding new SDK options (e.g. sendDefaultPii, profileSessionSampleRate), add them to the config type so they can be configured per environment.
#### Step 2: Import instrument.ts FIRST in src/main.ts
// instrument.ts MUST be the very first import — before NestJS or any other module
import "./instrument";
import { NestFactory } from "@nestjs/core";
import { AppModule } from "./app.module";
async function bootstrap() {
const app = await NestFactory.create(AppModule);
// Enable graceful shutdown — flushes Sentry events on SIGTERM/SIGINT
app.enableShutdownHooks();
await app.listen(3000);
}
bootstrap();
Why first? OpenTelemetry must monkey-patch http, express, database drivers, and other modules before they load. Any module that loads before instrument.ts will not be auto-instrumented.
#### Step 3: Register SentryModule and SentryGlobalFilter in src/app.module.ts
import { Module } from "@nestjs/common";
import { APP_FILTER } from "@nestjs/core";
import { SentryModule, SentryGlobalFilter } from "@sentry/nestjs/setup";
import { AppController } from "./app.controller";
import { AppService } from "./app.service";
@Module({
imports: [
SentryModule.forRoot(), // Registers SentryTracingInterceptor globally
],
controllers: [AppController],
providers: [
AppService,
{
provide: APP_FILTER,
useClass: SentryGlobalFilter, // Captures all unhandled exceptions
},
],
})
export class AppModule {}
What each piece does:
SentryModule.forRoot()— registersSentryTracingInterceptoras a globalAPP_INTERCEPTOR, enabling HTTP transaction naming
SentryGlobalFilter— extendsBaseExceptionFilter; captures exceptions across HTTP, GraphQL (rethrowsHttpExceptionwithout reporting), and RPC contexts
⚠️ **Do NOT register SentryModule.forRoot() twice.** If Phase 1 found it already imported in a shared library module (e.g. a SentryProxyModule or AnalyticsModule), do not add it again in AppModule. Duplicate registration causes every span to be intercepted twice, bloating trace data.
⚠️ Two entrypoints, different imports:
@sentry/nestjs→ SDK init, capture APIs, decorators (SentryTraced,SentryCron,SentryExceptionCaptured)
@sentry/nestjs/setup→ NestJS DI constructs (SentryModule,SentryGlobalFilter)
Never import SentryModule from @sentry/nestjs (main entrypoint) — it loads @nestjs/common before OpenTelemetry patches it, breaking auto-instrumentation.
ESM Setup (Node ≥ 18.19.0)
For ESM applications, use --import instead of a file import:
// instrument.mjs
import * as Sentry from "@sentry/nestjs";
Sentry.init({
dsn: process.env.SENTRY_DSN,
tracesSampleRate: 1.0,
});
// package.json
{
"scripts": {
"start": "node --import ./instrument.mjs -r ts-node/register src/main.ts"
}
}
Or via environment:
NODE_OPTIONS="--import ./instrument.mjs" npm run start
Exception Filter Options
Choose the approach that fits your existing architecture:
#### Option A: No existing global filter — use SentryGlobalFilter (recommended)
Already covered in Step 3 above. This is the simplest option.
#### Option B: Existing custom global filter — add @SentryExceptionCaptured() decorator
import { Catch, ExceptionFilter, ArgumentsHost } from "@nestjs/common";
import { SentryExceptionCaptured } from "@sentry/nestjs";
@Catch()
export class YourExistingFilter implements ExceptionFilter {
@SentryExceptionCaptured() // Wraps catch() to auto-report exceptions
catch(exception: unknown, host: ArgumentsHost): void {
// Your existing error handling continues unchanged
}
}
#### Option C: Specific exception type — manual capture
import { ArgumentsHost, Catch } from "@nestjs/common";
import { BaseExceptionFilter } from "@nestjs/core";
import * as Sentry from "@sentry/nestjs";
@Catch(ExampleException)
export class ExampleExceptionFilter extends BaseExceptionFilter {
catch(exception: ExampleException, host: ArgumentsHost) {
Sentry.captureException(exception);
super.catch(exception, host);
}
}
#### Option D: Microservice RPC exceptions
import { Catch, RpcExceptionFilter, ArgumentsHost } from "@nestjs/common";
import { Observable, throwError } from "rxjs";
import { RpcException } from "@nestjs/microservices";
import * as Sentry from "@sentry/nestjs";
@Catch(RpcException)
export class SentryRpcFilter implements RpcExceptionFilter<RpcException> {
catch(exception: RpcException, host: ArgumentsHost): Observable<any> {
Sentry.captureException(exception);
return throwError(() => exception.getError());
}
}
Decorators
#### @SentryTraced(op?) — Instrument any method
import { Injectable } from "@nestjs/common";
import { SentryTraced } from "@sentry/nestjs";
@Injectable()
export class OrderService {
@SentryTraced("order.process")
async processOrder(orderId: string): Promise<void> {
// Automatically wrapped in a Sentry span
}
@SentryTraced() // Defaults to op: "function"
async fetchInventory() { ... }
}
#### @SentryCron(slug, config?) — Monitor scheduled jobs
import { Injectable } from "@nestjs/common";
import { Cron } from "@nestjs/schedule";
import { SentryCron } from "@sentry/nestjs";
@Injectable()
export class ReportService {
@Cron("0 * * * *")
@SentryCron("hourly-report", {
// @SentryCron must come AFTER @Cron
schedule: { type: "crontab", value: "0 * * * *" },
checkinMargin: 2, // Minutes before marking missed
maxRuntime: 10, // Max runtime in minutes
timezone: "UTC",
})
async generateReport() {
// Check-in sent automatically on start/success/failure
}
}
#### Background Job Scope Isolation
Background jobs share the default isolation scope — wrap with Sentry.withIsolationScope() to prevent cross-contamination:
import * as Sentry from "@sentry/nestjs";
import { Injectable } from "@nestjs/common";
import { Cron, CronExpression } from "@nestjs/schedule";
@Injectable()
export class JobService {
@Cron(CronExpression.EVERY_HOUR)
handleCron() {
Sentry.withIsolationScope(() => {
Sentry.setTag("job", "hourly-sync");
this.doWork();
});
}
}
Apply withIsolationScope to: @Cron(), @Interval(), @OnEvent(), @Processor(), and any code outside the request lifecycle.
Working with Sentry DI Wrappers
Some NestJS projects wrap Sentry behind a dependency injection token (e.g. SENTRY_PROXY_TOKEN) for testability and decoupling. If Phase 1 detected this pattern, use the injected service for all runtime Sentry calls — do not import @sentry/nestjs directly in controllers, services, or processors.
import { Controller, Inject } from "@nestjs/common";
import { SENTRY_PROXY_TOKEN, type SentryProxyService } from "./sentry-proxy";
@Controller("orders")
export class OrderController {
constructor(
@Inject(SENTRY_PROXY_TOKEN) private readonly sentry: SentryProxyService,
private readonly orderService: OrderService,
) {}
@Post()
async createOrder(@Body() dto: CreateOrderDto) {
return this.sentry.startSpan(
{ name: "createOrder", op: "http" },
async () => this.orderService.create(dto),
);
}
}
**Where direct @sentry/nestjs import is still correct:**
instrument.ts— always usesimport * as Sentry from "@sentry/nestjs"forSentry.init()
- Standalone scripts and exception filters that run outside the DI container
Verification
Add a test endpoint to confirm events reach Sentry:
import { Controller, Get } from "@nestjs/common";
import * as Sentry from "@sentry/nestjs";
@Controller()
export class DebugController {
@Get("/debug-sentry")
triggerError() {
throw new Error("My first Sentry error from NestJS!");
}
@Get("/debug-sentry-span")
triggerSpan() {
return Sentry.startSpan({ op: "test", name: "NestJS Test Span" }, () => {
return { status: "span created" };
});
}
}
Hit GET /debug-sentry and check the Sentry Issues dashboard within seconds.
For Each Agreed Feature
Walk through features one at a time. Load the reference, follow its steps, verify before moving on:
Feature
Reference file
Load when...
Error Monitoring
${SKILL_ROOT}/references/error-monitoring.md
Always (baseline)
Tracing
${SKILL_ROOT}/references/tracing.md
Always (NestJS routes are auto-traced)
Profiling
${SKILL_ROOT}/references/profiling.md
CPU-intensive production apps
Logging
${SKILL_ROOT}/references/logging.md
Structured log aggregation needed
Metrics
${SKILL_ROOT}/references/metrics.md
Custom KPIs / SLO tracking
Crons
${SKILL_ROOT}/references/crons.md
Scheduled jobs or task queues
AI Monitoring
${SKILL_ROOT}/references/ai-monitoring.md
OpenAI/Anthropic/LangChain detected
For each feature: Read ${SKILL_ROOT}/references/<feature>.md, follow steps exactly, verify it works.
Configuration Reference
Key Sentry.init() Options
Option
Type
Default
Purpose
dsn
string
—
SDK disabled if empty; env: SENTRY_DSN
environment
string
"production"
e.g., "staging"; env: SENTRY_ENVIRONMENT
release
string
—
e.g., "myapp@1.0.0"; env: SENTRY_RELEASE
sendDefaultPii
boolean
false
Include IP addresses and request headers
tracesSampleRate
number
—
Transaction sample rate; undefined disables tracing
tracesSampler
function
—
Custom per-transaction sampling (overrides rate)
tracePropagationTargets
Array<string|RegExp>
—
URLs to propagate sentry-trace/baggage headers to
profileSessionSampleRate
number
—
Continuous profiling session rate (SDK ≥ 10.27.0)
profileLifecycle
"trace"|"manual"
"trace"
"trace" = auto-start profiler with spans; "manual" = call startProfiler()/stopProfiler()
enableLogs
boolean
false
Send structured logs to Sentry (SDK ≥ 9.41.0)
ignoreErrors
Array<string|RegExp>
[]
Error message patterns to suppress
ignoreTransactions
Array<string|RegExp>
[]
Transaction name patterns to suppress
beforeSend
function
—
Hook to mutate or drop error events
beforeSendTransaction
function
—
Hook to mutate or drop transaction events
beforeSendLog
function
—
Hook to mutate or drop log events
debug
boolean
false
Verbose SDK debug output
maxBreadcrumbs
number
100
Max breadcrumbs per event
Environment Variables
Variable
Maps to
Notes
SENTRY_DSN
dsn
Used if dsn not passed to init()
SENTRY_RELEASE
release
Also auto-detected from git SHA, Heroku, CircleCI
SENTRY_ENVIRONMENT
environment
Falls back to "production"
SENTRY_AUTH_TOKEN
CLI/source maps
For npx @sentry/wizard@latest -i sourcemaps
SENTRY_ORG
CLI/source maps
Organization slug
SENTRY_PROJECT
CLI/source maps
Project slug
Auto-Enabled Integrations
These integrations activate automatically when their packages are detected — no integrations: [...] needed:
Auto-enabled
Notes
httpIntegration
Outgoing HTTP calls via http/https/fetch
expressIntegration
Express adapter (default NestJS)
nestIntegration
NestJS lifecycle (middleware, guards, pipes, interceptors, handlers)
onUncaughtExceptionIntegration
Uncaught exceptions
onUnhandledRejectionIntegration
Unhandled promise rejections
openAIIntegration
OpenAI SDK (when installed)
anthropicAIIntegration
Anthropic SDK (when installed)
langchainIntegration
LangChain (when installed)
graphqlIntegration
GraphQL (when graphql package present)
postgresIntegration
pg driver
mysqlIntegration
mysql / mysql2
mongoIntegration
MongoDB / Mongoose
redisIntegration
ioredis / redis
Integrations Requiring Manual Setup
Integration
When to add
Code
nodeProfilingIntegration
Profiling desired
import { nodeProfilingIntegration } from "@sentry/profiling-node"
prismaIntegration
Prisma ORM used
integrations: [Sentry.prismaIntegration()]
consoleLoggingIntegration
Capture console output
integrations: [Sentry.consoleLoggingIntegration()]
localVariablesIntegration
Capture local var values in errors
integrations: [Sentry.localVariablesIntegration()]
Verification
Test that Sentry is receiving events:
// Add a test endpoint (remove before production)
@Get("/debug-sentry")
getError() {
throw new Error("My first Sentry error!");
}
Or send a test message without crashing:
import * as Sentry from "@sentry/nestjs";
Sentry.captureMessage("NestJS Sentry SDK test");
If nothing appears:
- Set
debug: trueinSentry.init()— prints SDK internals to stdout
- Verify
SENTRY_DSNenv var is set in the running process
- Check that
import "./instrument"is the first line inmain.ts
- Confirm
SentryModule.forRoot()is imported inAppModule
- Check DSN format:
https://<key>@o<org>.ingest.sentry.io/<project>
Phase 4: Cross-Link
After completing NestJS setup, check for a companion frontend missing Sentry:
ls -d ../frontend ../web ../client ../ui 2>/dev/null
cat ../frontend/package.json ../web/package.json 2>/dev/null \
| grep -E '"react"|"svelte"|"vue"|"next"|"nuxt"'
If a frontend exists without Sentry, suggest the matching skill:
Frontend detected
Suggest skill
Next.js
sentry-nextjs-sdk
React
sentry-react-sdk
Svelte / SvelteKit
sentry-svelte-sdk
Vue / Nuxt
Use @sentry/vue — see docs.sentry.io/platforms/javascript/guides/vue/
React Native / Expo
sentry-react-native-sdk
Troubleshooting
Issue
Solution
Events not appearing
Set debug: true, verify SENTRY_DSN, check instrument.ts is imported first
Malformed DSN error
Format: https://<key>@o<org>.ingest.sentry.io/<project>
Exceptions not captured
Ensure SentryGlobalFilter is registered via APP_FILTER in AppModule
Auto-instrumentation not working
instrument.ts must be the first import in main.ts — before all NestJS imports
Profiling not starting
Requires tracesSampleRate > 0 + profileSessionSampleRate > 0 + @sentry/profiling-node installed
enableLogs not working
Requires SDK ≥ 9.41.0
No traces appearing
Verify tracesSampleRate is set (not undefined)
Too many transactions
Lower tracesSampleRate or use tracesSampler to drop health checks
Fastify + GraphQL issues
Known edge cases — see GitHub #13388; prefer Express for GraphQL
Background job events mixed
Wrap job body in Sentry.withIsolationScope(() => { ... })
Prisma spans missing
Add integrations: [Sentry.prismaIntegration()] to Sentry.init()
ESM syntax errors
Set registerEsmLoaderHooks: false (disables ESM hooks; also disables auto-instrumentation for ESM modules)
SentryModule breaks instrumentation
Must import from @sentry/nestjs/setup, never from @sentry/nestjs
RPC exceptions not captured
Add dedicated SentryRpcExceptionFilter (see Option D in exception filter section)
WebSocket exceptions not captured
Use @SentryExceptionCaptured() on gateway handleConnection/handleDisconnect
@SentryCron not triggering
Decorator order matters — @SentryCron MUST come after @Cron
TypeScript path alias issues
Ensure tsconfig.json paths are configured so instrument resolves from main.ts location
import * as Sentry ESLint error
Many projects ban namespace imports. Use named imports (import { startSpan, captureException } from "@sentry/nestjs") or use the project's DI proxy instead
profilesSampleRate vs profileSessionSampleRate
profilesSampleRate is deprecated in SDK 10.x. Use profileSessionSampleRate + profileLifecycle: "trace" instead
Duplicate spans on every request
SentryModule.forRoot() registered in multiple modules. Ensure it's only called once — check shared/library modules
Config property not recognized in instrument.ts
When using a typed config class, new SDK options must be added to the config type definition and the project rebuilt before TypeScript recognizes them
Version Requirements
Feature
Minimum SDK Version
@sentry/nestjs package
8.0.0
@SentryTraced decorator
8.15.0
@SentryCron decorator
8.16.0
Event Emitter auto-instrumentation
8.39.0
SentryGlobalFilter (unified)
8.40.0
Sentry.logger API (enableLogs)
9.41.0
profileSessionSampleRate
10.27.0
Node.js requirement
≥ 18
Node.js for ESM --import
≥ 18.19.0
NestJS compatibility
8.x – 11.x