SKILL.md
$27
Phase 1: Detect
Run these commands to understand the project before making any recommendations:
# Detect Cloudflare project
ls wrangler.toml wrangler.jsonc wrangler.json 2>/dev/null
# Detect existing Sentry
cat package.json 2>/dev/null | grep -E '"@sentry/'
# Detect project type (Workers vs Pages)
ls functions/ functions/_middleware.js functions/_middleware.ts 2>/dev/null && echo "Pages detected"
cat wrangler.toml 2>/dev/null | grep -E 'main|pages_build_output_dir'
# Detect framework
cat package.json 2>/dev/null | grep -E '"hono"|"remix"|"astro"|"svelte"'
# Detect Durable Objects
cat wrangler.toml 2>/dev/null | grep -i 'durable_objects'
# Detect D1 databases
cat wrangler.toml 2>/dev/null | grep -i 'd1_databases'
# Detect Queues
cat wrangler.toml 2>/dev/null | grep -i 'queues'
# Detect Workflows
cat wrangler.toml 2>/dev/null | grep -i 'workflows'
# Detect Scheduled handlers (cron triggers)
cat wrangler.toml 2>/dev/null | grep -i 'crons\|triggers'
# Detect compatibility flags
cat wrangler.toml 2>/dev/null | grep -i 'compatibility_flags'
cat wrangler.jsonc 2>/dev/null | grep -i 'compatibility_flags'
# Detect AI/LLM libraries
cat package.json 2>/dev/null | grep -E '"openai"|"@anthropic-ai"|"ai"|"@google/generative-ai"|"@langchain"'
# Detect logging libraries
cat package.json 2>/dev/null | grep -E '"pino"|"winston"'
# Check for companion frontend
ls frontend/ web/ client/ 2>/dev/null
cat package.json 2>/dev/null | grep -E '"react"|"vue"|"svelte"|"next"'
What to determine:
Question
Impact
Workers or Pages?
Determines wrapper: withSentry vs sentryPagesPlugin
Hono framework?
Automatic Hono error handler integration via honoIntegration
@sentry/cloudflare already installed?
Skip install, go to feature config
Durable Objects configured?
Recommend instrumentDurableObjectWithSentry
D1 databases bound?
Recommend instrumentD1WithSentry
Queues configured?
withSentry auto-instruments queue handlers
Workflows configured?
Recommend instrumentWorkflowWithSentry
Cron triggers configured?
withSentry auto-instruments scheduled handlers; recommend Crons monitoring
nodejs_als or nodejs_compat flag set?
Required — SDK needs AsyncLocalStorage
AI/LLM libraries?
Recommend AI Monitoring integrations
Companion frontend?
Trigger Phase 4 cross-link
Phase 2: Recommend
Present a concrete recommendation based on what you found. Don't ask open-ended questions — lead with a proposal:
Recommended (core coverage):
- ✅ Error Monitoring — always; captures unhandled exceptions in fetch, scheduled, queue, email, and Durable Object handlers
- ✅ Tracing — automatic HTTP request spans, outbound fetch tracing, D1 query spans
Optional (enhanced observability):
- ⚡ Logging — structured logs via
Sentry.logger.*; recommend when log search is needed
- ⚡ Crons — detect missed/failed scheduled jobs; recommend when cron triggers are configured
- ⚡ D1 Instrumentation — automatic query spans and breadcrumbs; recommend when D1 is bound
- ⚡ Durable Objects — automatic error capture and spans for DO methods; recommend when DOs are configured
- ⚡ Workflows — automatic span creation for workflow steps; recommend when Workflows are configured
- ⚡ AI Monitoring — Vercel AI SDK, OpenAI, Anthropic, LangChain; recommend when AI libraries detected
Recommendation logic:
Feature
Recommend when...
Error Monitoring
Always — non-negotiable baseline
Tracing
Always — HTTP request tracing and outbound fetch are high-value
Logging
App needs structured log search or log-to-trace correlation
Crons
Cron triggers configured in wrangler.toml
D1 Instrumentation
D1 database bindings present
Durable Objects
Durable Object bindings configured
Workflows
Workflow bindings configured
AI Monitoring
App uses Vercel AI SDK, OpenAI, Anthropic, or LangChain
Metrics
App needs custom counters, gauges, or distributions
Propose: "I recommend setting up Error Monitoring + Tracing. Want me to also add D1 instrumentation and Crons monitoring?"
Phase 3: Guide
Option 1: Source Maps Wizard
You need to run this yourself — the wizard opens a browser for login and requires interactive input that the agent can't handle. Copy-paste into your terminal:
npx @sentry/wizard@latest -i sourcemaps
This sets up source map uploading so your production stack traces show readable code. It does not set up the SDK initialization — you still need to follow Option 2 below for the actual SDK setup.
Once it finishes, continue with Option 2 for SDK setup.
Note: Unlike framework SDKs (Next.js, SvelteKit), there is no Cloudflare-specific wizard integration. The sourcemaps wizard only handles source map upload configuration.
Option 2: Manual Setup
#### Prerequisites: Compatibility Flags
The SDK requires AsyncLocalStorage. Add one of these flags to your Wrangler config:
wrangler.toml:
compatibility_flags = ["nodejs_als"]
# or: compatibility_flags = ["nodejs_compat"]
wrangler.jsonc:
{
"compatibility_flags": ["nodejs_als"]
}
nodejs_als is lighter — it only enables AsyncLocalStorage. Use nodejs_compat if your code also needs other Node.js APIs.
#### Install
npm install @sentry/cloudflare
#### Workers Setup
Wrap your handler with withSentry. This automatically instruments fetch, scheduled, queue, email, and tail handlers:
import * as Sentry from "@sentry/cloudflare";
export default Sentry.withSentry(
(env: Env) => ({
dsn: env.SENTRY_DSN,
sendDefaultPii: true,
tracesSampleRate: 1.0,
enableLogs: true,
}),
{
async fetch(request, env, ctx) {
return new Response("Hello World!");
},
} satisfies ExportedHandler<Env>,
);
Key points:
- The first argument is a callback that receives
env— use this to read secrets likeSENTRY_DSN
- The SDK reads DSN, environment, release, debug, tunnel, and traces sample rate from
envautomatically (see [Environment Variables](#environment-variables))
withSentrywraps all exported handlers — you do not need separate wrappers forscheduled,queue, etc.
#### Pages Setup
Use sentryPagesPlugin as middleware:
// functions/_middleware.ts
import * as Sentry from "@sentry/cloudflare";
export const onRequest = Sentry.sentryPagesPlugin((context) => ({
dsn: context.env.SENTRY_DSN,
sendDefaultPii: true,
tracesSampleRate: 1.0,
enableLogs: true,
}));
Chaining multiple middlewares:
import * as Sentry from "@sentry/cloudflare";
export const onRequest = [
// Sentry must be first
Sentry.sentryPagesPlugin((context) => ({
dsn: context.env.SENTRY_DSN,
tracesSampleRate: 1.0,
})),
// Add more middlewares here
];
**Using wrapRequestHandler directly** (for frameworks like SvelteKit on Cloudflare Pages):
import * as Sentry from "@sentry/cloudflare";
export const handle = ({ event, resolve }) => {
return Sentry.wrapRequestHandler(
{
options: {
dsn: event.platform.env.SENTRY_DSN,
tracesSampleRate: 1.0,
},
request: event.request,
context: event.platform.ctx,
},
() => resolve(event),
);
};
#### Hono on Cloudflare Workers
Hono apps are objects with a fetch method — wrap them with withSentry directly:
import { Hono } from "hono";
import * as Sentry from "@sentry/cloudflare";
const app = new Hono();
app.get("/", (ctx) => ctx.json({ message: "Hello" }));
app.get("/error", () => {
throw new Error("Test error");
});
app.onError((err, ctx) => {
return ctx.json({ error: err.message }, 500);
});
export default Sentry.withSentry(
(env: Env) => ({
dsn: env.SENTRY_DSN,
tracesSampleRate: 1.0,
}),
app,
);
The honoIntegration (enabled by default) automatically captures errors from Hono's onError handler and sets the correct transaction name with the route path.
#### Set Up the SENTRY_DSN Secret
Store your DSN as a Cloudflare secret — do not hardcode it:
# Local development: add to .dev.vars
echo 'SENTRY_DSN="https://examplePublicKey@o0.ingest.sentry.io/0"' >> .dev.vars
# Production: set as a secret
npx wrangler secret put SENTRY_DSN
Add the binding to your Env type:
interface Env {
SENTRY_DSN: string;
// ... other bindings
}
#### Source Maps Setup
Source maps make production stack traces readable. Without them, you see minified/bundled code.
Step 1: Generate a Sentry auth token
Go to sentry.io/settings/auth-tokens/ and create a token with project:releases and org:read scopes.
Step 2: Install the Sentry Vite plugin (most Cloudflare projects use Vite via Wrangler):
npm install @sentry/vite-plugin --save-dev
**Step 3: Configure vite.config.ts** (if your project has one):
import { defineConfig } from "vite";
import { sentryVitePlugin } from "@sentry/vite-plugin";
export default defineConfig({
build: {
sourcemap: true,
},
plugins: [
sentryVitePlugin({
org: "___ORG_SLUG___",
project: "___PROJECT_SLUG___",
authToken: process.env.SENTRY_AUTH_TOKEN,
}),
],
});
Step 4: Set environment variables in CI
SENTRY_AUTH_TOKEN=sntrys_eyJ...
SENTRY_ORG=my-org
SENTRY_PROJECT=my-project
**Step 5: Add to .gitignore**
.dev.vars
.env.sentry-build-plugin
Automatic Release Detection
The SDK can automatically detect the release version via Cloudflare's version metadata binding:
wrangler.toml:
[version_metadata]
binding = "CF_VERSION_METADATA"
Release priority (highest to lowest):
releaseoption passed toSentry.init()
SENTRY_RELEASEenvironment variable
CF_VERSION_METADATA.idbinding
For Each Agreed Feature
Load the corresponding reference file and follow its steps:
Feature
Reference file
Load when...
Error Monitoring
references/error-monitoring.md
Always (baseline) — unhandled exceptions, manual capture, scopes, enrichment
Tracing
references/tracing.md
HTTP request tracing, outbound fetch spans, D1 query spans, distributed tracing
Logging
references/logging.md
Structured logs via Sentry.logger.*, log-to-trace correlation
Crons
references/crons.md
Scheduled handler monitoring, withMonitor, check-in API
Durable Objects
references/durable-objects.md
Instrument Durable Object classes for error capture and spans
For each feature: read the reference file, follow its steps exactly, and verify before moving on.
Verification
After setup, verify Sentry is working:
// Add temporarily to your fetch handler, then remove
export default Sentry.withSentry(
(env: Env) => ({
dsn: env.SENTRY_DSN,
tracesSampleRate: 1.0,
}),
{
async fetch(request, env, ctx) {
throw new Error("Sentry test error — delete me");
},
} satisfies ExportedHandler<Env>,
);
Deploy and trigger the route, then check your Sentry Issues dashboard — the error should appear within ~30 seconds.
Verification checklist:
Check
How
Errors captured
Throw in a fetch handler, verify in Sentry
Tracing working
Check Performance tab for HTTP spans
Source maps working
Check stack trace shows readable file/line names
D1 spans (if configured)
Run a D1 query, check for db.query spans
Scheduled monitoring (if configured)
Trigger a cron, check Crons dashboard
Config Reference
Sentry.init() Options
Option
Type
Default
Notes
dsn
string
—
Required. Read from env.SENTRY_DSN automatically if not set
tracesSampleRate
number
—
0–1; 1.0 in dev, lower in prod recommended
tracesSampler
function
—
Dynamic sampling function; mutually exclusive with tracesSampleRate
sendDefaultPii
boolean
false
Include request headers and cookies in events
enableLogs
boolean
false
Enable Sentry Logs product
environment
string
auto
Read from env.SENTRY_ENVIRONMENT if not set
release
string
auto
Detected from CF_VERSION_METADATA.id or SENTRY_RELEASE
debug
boolean
false
Read from env.SENTRY_DEBUG if not set. Log SDK activity to console
tunnel
string
—
Read from env.SENTRY_TUNNEL if not set
beforeSend
function
—
Filter/modify error events before sending
beforeSendTransaction
function
—
Filter/modify transaction events before sending
beforeSendLog
function
—
Filter/modify log entries before sending
tracePropagationTargets
(string|RegExp)[]
all URLs
Control which outbound requests get trace headers
skipOpenTelemetrySetup
boolean
false
Opt-out of OpenTelemetry compatibility tracer
instrumentPrototypeMethods
boolean | string[]
false
Durable Object: instrument prototype methods for RPC spans
Environment Variables (Read from env )
The SDK reads these from the Cloudflare env object automatically:
Variable
Purpose
SENTRY_DSN
DSN for Sentry init
SENTRY_RELEASE
Release version string
SENTRY_ENVIRONMENT
Environment name (production, staging)
SENTRY_TRACES_SAMPLE_RATE
Traces sample rate (parsed as float)
SENTRY_DEBUG
Enable debug mode ("true" / "1")
SENTRY_TUNNEL
Tunnel URL for event proxying
CF_VERSION_METADATA
Cloudflare version metadata binding (auto-detected release)
Default Integrations
These are registered automatically by getDefaultIntegrations():
Integration
Purpose
dedupeIntegration
Prevent duplicate events (disabled for Workflows)
inboundFiltersIntegration
Filter events by type, message, URL
functionToStringIntegration
Preserve original function names
linkedErrorsIntegration
Follow cause chains in errors
fetchIntegration
Trace outbound fetch() calls, create breadcrumbs
honoIntegration
Auto-capture Hono onError exceptions
requestDataIntegration
Attach request data to events
consoleIntegration
Capture console.* calls as breadcrumbs
Phase 4: Cross-Link
After completing Cloudflare setup, check for companion services:
# Check for companion frontend
ls frontend/ web/ client/ ui/ 2>/dev/null
cat package.json 2>/dev/null | grep -E '"react"|"vue"|"svelte"|"next"|"astro"'
# Check for companion backend in adjacent directories
ls ../backend ../server ../api 2>/dev/null
cat ../go.mod ../requirements.txt ../Gemfile 2>/dev/null | head -3
If a frontend is found, suggest the matching SDK skill:
Frontend detected
Suggest skill
React
sentry-react-sdk
Next.js
sentry-nextjs-sdk
Svelte/SvelteKit
sentry-svelte-sdk
Vue/Nuxt
See docs.sentry.io/platforms/javascript/guides/vue/
If a backend is found in a different directory:
Backend detected
Suggest skill
Go (go.mod)
sentry-go-sdk
Python (requirements.txt, pyproject.toml)
sentry-python-sdk
Ruby (Gemfile)
sentry-ruby-sdk
Node.js (Express, Fastify)
sentry-node-sdk
Connecting frontend and backend with linked Sentry projects enables distributed tracing — stack traces that span your browser, Cloudflare Worker, and backend API in a single trace view.
Troubleshooting
Issue
Cause
Solution
Events not appearing
DSN not set or debug: false hiding errors
Set debug: true temporarily in init options; verify SENTRY_DSN secret is set with wrangler secret list
AsyncLocalStorage is not defined
Missing compatibility flag
Add nodejs_als or nodejs_compat to compatibility_flags in wrangler.toml
Stack traces show minified code
Source maps not uploaded
Configure @sentry/vite-plugin or run npx @sentry/wizard -i sourcemaps; verify SENTRY_AUTH_TOKEN in CI
Events lost on short-lived requests
SDK not flushing before worker terminates
Ensure withSentry or sentryPagesPlugin wraps your handler — they use ctx.waitUntil() to flush
Hono errors not captured
Hono app not wrapped with withSentry
Pass the Hono app as the second argument to Sentry.withSentry()
Durable Object errors missing
DO class not instrumented
Wrap class with Sentry.instrumentDurableObjectWithSentry() — see references/durable-objects.md
D1 queries not creating spans
D1 binding not instrumented
Wrap binding with Sentry.instrumentD1WithSentry(env.DB) before use
Scheduled handler not monitored
withSentry not wrapping the handler
Ensure export default Sentry.withSentry(...) wraps your entire exported handler object
Release not auto-detected
CF_VERSION_METADATA binding not configured
Add [version_metadata] with binding = "CF_VERSION_METADATA" to wrangler.toml
Duplicate events in Workflows
Dedupe integration filtering step failures
SDK automatically disables dedupe for Workflows; verify you use instrumentWorkflowWithSentry