Global State Ownership Map
Write down which store owns each piece of shared state so nothing lives in two places.
What it adds
A written map of every shared state domain in the app and the single store, cache, or provider that owns it.
What your agent is told to do
5
What your agent is told to do
5-
1
Inventory the shared state the app already holds — the signed-in user, the active workspace, the theme, the cart, the open record, the notification count — and name the one place each is read from.
-
2
Split the map into two columns: state that mirrors something the server owns, and state that only exists in the browser. The first is a cache with a refresh and an invalidation story; the second is genuinely local and needs a reset story instead.
-
3
For every server-derived entry, record what invalidates it and what clears it on sign-out or workspace switch. An entry with no documented invalidation is a bug waiting for a stale price or a stale permission.
-
4
Where the same value is currently held in more than one place, pick the owner, and make the other places read from it rather than keep their own copy. Deleting the duplicate is the point of the exercise.
-
5
Do not treat this as a rewrite. The deliverable is the map plus the removal of duplicated ownership, not a migration to a different state library — swapping the library leaves the same value in three places under new names.
Edge cases it handles
7
Edge cases it handles
7- A value held in two stores will drift, and the drift shows up as a price, a permission, or an unread count that disagrees with itself between two panels on the same screen. Every domain needs exactly one owner named in the map.
- Server cache and client state have different lifetimes and must not share a container. A cached list can be thrown away and refetched at any moment; a half-typed filter panel cannot, and putting them together means one policy is wrong for one of them.
- Every entry needs both an invalidation rule and a reset rule, and they are not the same thing. Invalidation says when the cached copy is no longer trustworthy; reset says what happens on sign-out, workspace switch, or an impersonation session ending.
- Lazily loaded parts of the app must be able to join the map without a circular import back to the code that loads them. If a store can only be reached from the root bundle, code splitting quietly pulls the whole thing back in.
- State restored from storage on boot must be validated against the current shape before it is trusted, or a stale entry from an older release will be handed to code that no longer understands it.
- Two tabs open on the same account share persisted state but not in-memory state, and the map must say which entries are expected to converge across tabs and which are deliberately per-tab.
- Anything holding a token, a session, or personal data must be marked in the map so it is obvious what has to be cleared on sign-out.
Definition of done
8
Definition of done
8- Every shared state domain in the app appears in the map with exactly one named owner.
- Server-derived caches and browser-only state are listed separately and stored separately.
- Each cached domain documents what invalidates it and what clears it on sign-out or workspace switch.
- No value is held authoritatively in two stores; duplicated copies have been removed rather than kept in sync.
- Lazily loaded areas register their state without forcing their code into the initial bundle.
- Entries holding credentials or personal data are marked, and all of them are cleared on sign-out.
- 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
Prompt Versioning
Prompt Versioning
Tie every AI output to the exact prompt version that produced it.
What it does
Immutable, numbered versions of each prompt, with the run configuration recorded and every output stamped with the version used.
How it works
- 1 Make every publish create a new immutable version rather than overwriting the previous text. Editing history in place destroys the only record of what produced last month's outputs.
- 2 Capture the whole run configuration with each version, not just the wording: which model tier and parameters were used, which tools were available, and the expected output shape. A prompt that behaves differently under different settings is not one prompt.
- 3 Stamp every generated output with the version identifier that produced it, and keep that stamp with the record so an output found later can be traced back to its exact instructions.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/prompt-versioning
Retrieval Debugger
Retrieval Debugger
Show exactly which sources, chunks, and scores produced a given AI answer.
What it does
A per-answer inspector showing the query as issued, the filters applied, the candidate chunks with their scores, and what reached the model.
How it works
- 1 Capture for each answer the query as it was issued, the filters applied, the candidates returned with their scores, and which of those actually made it into the request after the context ceiling was applied.
- 2 Show results after permission filtering, with a count of how many candidates were excluded and why. Displaying the pre-filter set turns the debugger into a way to read content the viewer cannot open.
- 3 Present each scoring stage separately — keyword, semantic, and any reranking — because a chunk that ends up first overall may have been rescued by one stage after being buried by another, and a single blended number hides that.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/retrieval-debugger
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.