Read and Unread State
Show people which documents, threads, and records they have not looked at yet.
What it adds
Per-user read state on content items — documents, records, comment threads, list rows — with unread markers and counts that stay correct across devices and bulk actions.
What your agent is told to do
6
What your agent is told to do
6-
1
Apply this to content the user browses, not to notifications. If the app already has a notification inbox with unread handling, reuse that state machine for content rather than forking a second, subtly different one.
-
2
Decide and document the single rule for what marks an item read — opening it, dwelling on it for a moment, or an explicit action. Pick one and apply it everywhere.
-
3
Store read state per user, on the server. Client-only state means a phone and a laptop disagree about the same inbox.
-
4
Give the user manual control in both directions: mark as read, mark as unread, and mark all as read.
-
5
Derive unread counts from the same query that renders the list, so the badge and the list can never disagree.
-
6
Do NOT mark items read simply because they scrolled past. Passive scroll marking silently buries things people meant to come back to.
Edge cases it handles
7
Edge cases it handles
7- Deleting or archiving an unread item must decrement the count immediately, not leave a badge pointing at nothing.
- Unread counts must respect the active filter or scope, and the UI must be clear about which scope the number describes.
- Bulk mark-as-read must be undoable, or at least confirmed, when it covers more than a screenful.
- Re-importing or reprocessing data must not resurrect items as unread. Key read state to the item, not to the row version.
- Two open tabs must converge. Marking read in one should not leave the other showing a stale badge indefinitely.
- Cap the displayed count at a sensible ceiling (99+) rather than rendering a four-digit badge that breaks the layout.
- In shared or team contexts, be explicit about whether read state is personal or shared. Guessing wrong here is a trust problem.
Definition of done
8
Definition of done
8- Read state is stored per user on the server and matches across devices.
- One documented rule decides when an item becomes read.
- Users can mark items read and unread individually, and read in bulk.
- Unread counts stay correct after deletion, archiving, filtering, and bulk actions.
- Reimported data does not flip existing items back to unread.
- The badge count and the visible list never contradict each other.
- The feature matches the existing design system.
- No existing functionality is broken.
Related features
Poll Widget
Poll Widget
Ask a one-question poll anywhere in the app and show the results as soon as someone votes.
What it does
An embeddable single-question poll with a vote control and a results view revealed after voting.
How it works
- 1 Keep it to one question with a small set of options. The moment it grows a second question it is a survey and belongs in the survey feature instead.
- 2 Record the vote on the server and return the updated tallies in the same response, so the results the voter sees are the real counts rather than an optimistic local guess that may disagree after a refresh.
- 3 Hold voting open to signed-out visitors if the poll is public, and identify them well enough to prevent casual repeat voting without demanding an account for a single click.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/poll-widget
Discussion Forum
Discussion Forum
Give users a public place to ask questions and answer each other.
What it does
Topic boards containing user-created threads with replies, moderation controls, and permalinks.
How it works
- 1 Ship boards, threads, replies, and a single level of nesting. Resist arbitrary-depth reply trees, reputation scores, badges, and private subforums until people are actually posting; each of those is a feature in its own right.
- 2 Reuse Comments and Mentions for the reply composer, mention parsing, and content rendering rather than writing a second editor with its own escaping rules.
- 3 Route reported and held posts into the existing Moderation Queue and reuse AI Text Content Moderation and Content Reporting if present. Do not build a forum-specific moderation inbox alongside the one the app already has.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/discussion-forum
Prize Wheel Giveaway
Prize Wheel Giveaway
Run a spin-the-wheel giveaway where each entrant gets one spin and a recorded prize.
What it does
A giveaway with a wheel interface, one server-decided spin per entrant, and a durable record of every result.
How it works
- 1 Decide the outcome on the server before the wheel starts turning. The request returns a prize, the animation lands on it, and the client is only ever playing back a result it was given.
- 2 Model prizes with explicit weights and optional stock counts, and let an operator edit them without a deploy. A giveaway whose odds live in code cannot be corrected once it is live.
- 3 Tie an entry to an account where the app has one, and where it does not, combine the strongest identifiers available and treat a single spin as the hard limit across all of them.
Copy the prompt
No account needed
Add this feature to my app:
https://addthisfeature.com/x/prize-wheel-giveaway
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.