Sentry Error Forwarding
Send real application errors to Sentry with clean context and grouping worth acting on.
What it adds
Server and client error reporting into Sentry, with release and environment tagging, scrubbing, and deliberate noise control.
What your agent is told to do
5
What your agent is told to do
5-
1
Find the app's existing error handling paths and route from there. Reporting belongs where errors are already caught and logged, not sprinkled into individual controllers and components.
-
2
Separate genuine faults from expected outcomes. Validation failures, permission denials, not-found responses, and cancelled requests are normal behaviour and must not be reported as crashes, or the signal is buried within a day.
-
3
Scrub every event before it leaves the process: authorisation headers, tokens, passwords, payment details, request bodies, query strings, and any personal data the app is not permitted to send to a third party. Attach an internal user or account identifier instead of a name or an email address.
-
4
Tag every event with the deployed release and the environment, taken from the same source the deploy pipeline already uses, so a spike can be traced to a specific ship. Keep the credentials for the reporting endpoint in server configuration and use the separate client-side key for browser reporting.
-
5
Leave grouping to the provider's default and override the fingerprint only for the specific cases where it is demonstrably wrong — a wrapper exception that collapses thousands of distinct faults into one, or a message carrying an identifier that splits one fault into thousands.
Edge cases it handles
8
Edge cases it handles
8- Events routinely carry secrets and personal data by accident, through request headers, form bodies, and exception messages that interpolate a value. Scrub at the point of capture, and default to dropping unknown fields rather than sending them.
- A custom fingerprint applied broadly is worse than the default, because it merges unrelated faults into a single unactionable issue. Set one only where the default grouping has been shown to fail, and record why.
- Release and environment must be set identically everywhere, including background jobs and the browser build. A missing release makes a regression impossible to attribute to a deploy.
- Expected errors reported as crashes destroy the error budget and train the team to ignore alerts. Classify them explicitly and let them stay in ordinary logs.
- Linking an error to a trace is useful, but reporting the same fault at every layer it bubbles through floods the project with duplicates. Report once, at the outermost handler, and carry the trace identifier rather than re-raising a new event per frame.
- A single failing endpoint can generate an enormous burst and exhaust the account's quota within minutes. Sample repetitive events and rate-limit reporting per issue rather than sending everything.
- The reporting service itself will be unavailable or slow. Sending must never block a user request or fail one; drop or queue the event and carry on.
- Errors thrown during startup, or inside the reporting path itself, must not recurse. Guard the reporter so a failure to report cannot generate another report.
Definition of done
8
Definition of done
8- Errors are reported from the app's existing central handlers rather than from scattered call sites.
- Expected validation, permission, and not-found outcomes do not appear as reported errors.
- No secret, credential, or personal data field is present in any forwarded event, verified against a real captured payload.
- Every event carries a release and an environment consistent with the deploy pipeline, in both server and browser reporting.
- Custom fingerprints exist only where default grouping was inadequate, and each one is documented.
- A burst of identical errors is sampled and rate-limited, and reporting never blocks or fails a user request.
- The feature matches the existing design system.
- No existing functionality is broken.
Related features
AI Cost Budgets
AI Cost Budgets
Cap what AI features are allowed to spend before the bill arrives.
What it does
Monetary spending limits on AI work, scoped by workspace, feature, and time period, enforced before a run starts.
How it works
- 1 Find every place the app calls a model and route all of them through one accounting point that records estimated and actual spend against a scope. A budget that only covers the chat feature is not a budget.
- 2 Estimate the cost of a run from the size of its input before dispatching it, and refuse anything that would exceed the remaining budget on its own.
- 3 Reserve the estimate against the budget when the run starts, then reconcile to the real usage figures when it finishes, releasing whatever was over-reserved.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/ai-cost-budgets
Multi-Model Routing
Multi-Model Routing
Send each AI request to the right model using rules you can read and test.
What it does
A deterministic routing layer that picks a model per request from task type, context size, latency budget, and data sensitivity.
How it works
- 1 Express routing as explicit, ordered rules over inputs the app can measure: task type, estimated context size, latency budget, and the sensitivity classification of the data involved. A rule set that can be read line by line can be reviewed and tested.
- 2 Make routing deterministic. The same inputs must always produce the same route, so a bad output can be reproduced and a rule change can be evaluated. Randomised or load-based selection turns every incident into guesswork.
- 3 Classify data before routing and refuse to route restricted content to any destination not approved for it. This check is a hard block, not a preference, and it must run before the request is assembled.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/multi-model-routing
Model Selection
Model Selection
Let each AI task run on the model that suits its quality, speed, and cost needs.
What it does
A per-task model choice, drawn from the models the app already has configured, with capability filtering and safe defaults.
How it works
- 1 Enumerate the models the app already has access to and record what each one can actually do: context capacity, whether it can return the structured output the task requires, whether it supports the tools the task calls, and its relative cost and speed.
- 2 Offer only the models that satisfy the task's requirements. A task that needs structured output must not list a model that cannot reliably produce it, because the failure appears later as malformed responses rather than as an unavailable option.
- 3 Store the choice against the specific task, not as one global setting. A single default forces a summarisation task and a classification task onto the same tier when they have opposite needs.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/model-selection
How it works
-
1
Copy the link
Grab the Markdown instruction URL for this feature.
-
2
Give it to your AI
Paste it into Claude Code, Cursor, v0, Lovable — whatever you build with.
-
3
It inspects, then implements
Your agent reads your existing app first, then adds the feature to fit it.
Works with your stack
These instructions are written to adapt. They tell the agent to detect your framework, match your existing design system, and reuse what you already have — rather than assuming a particular stack.
Need it tighter than that? Customize the feature and tell it exactly what you're running.