Privacy
This tool refuses to collect or store personal information. Not “we handle it carefully” — we don’t take it in the first place. The safest place for your data is somewhere it was never collected. Where a model has to see your words to coach you, it keeps none of them.
What this site doesn’t collect
Accounts and passwords
There is nothing to sign up for on this site, so there is no user record to hold, breach or delete.
Analytics and tracking
No analytics scripts, no advertising or tracking cookies, no fingerprinting. We don’t know who visited, or which pages you read.
Contact and mailing details
No newsletter, no “enter your email to download” — every template and skill downloads straight away. The feedback form asks for contact details only as an optional field you can leave blank.
Anything you didn’t press send on
Content is markdown files in the repository; ideas are GitHub issues. The one database here holds feedback people chose to send us, and nothing else — no visits, no page views, no profile built from either.
Pages are served by Vercel, which keeps short-lived operational request logs (including IP addresses) to run and protect the servers, as any web host does. We don’t read them for analytics and we add nothing to them.
When you contribute
Ideas, comments and endorsements live as GitHub issues, under your own GitHub account and GitHub’s privacy terms — not ours. That means everything you contribute is public, attributed to your GitHub name or display name, and yours to edit or delete. We hold no copy of your profile.
The ideas board is a separate app. Moving a card asks for a shared facilitator passcode, held in a cookie in your own browser for 12 hours — it identifies the workshop, not you.
When you send feedback
The feedback form is the one thing on this site that writes anything down, and only when you ask it to. Your answers travel in the page address so the page can show you a summary first, and nothing leaves the page until you press send. Every question is optional, none of them asks who you are, and the contact field is yours to leave blank.
Then you pick where it goes. Sending it to us stores one row in our database: the summary you just read, the contact details only if you typed some, and the time. Nothing else about the visit is stored with it — no IP address, no browser, no cookie, no identifier that would let two submissions be tied to the same person — so there is no profile here to build, breach or hand over. Posting it on GitHub sends it publicly from your own browser, under your GitHub name, where you can edit or delete it yourself. Copying the summary and pasting it elsewhere sends it with no identity attached at all. Ask us and we will delete anything you sent.
When the coach helps with a goal
The Outcome Coach on this site sends the goal you paste, and any answers you give it, to an AI model through Vercel’s AI Gateway to write the review. That is the one place text you type leaves this site. We store none of it: there is no account, nothing about it is written to our database, and the conversation exists only on the page in front of you until you leave it. Neither does the model — every call runs under zero data retention. Anonymise anything sensitive anyway — the coaching is just as good on a redacted version, and a habit beats a promise. If the AI coach isn’t switched on for a deployment, the page runs a structural check that sends nothing anywhere.
The coaching skills you download run inside your own AI assistant — nothing comes back here. Both the on-site coach and the skills are written to refuse personal information: they never ask for names, contact details, or individual performance or health data, and if you paste some anyway they say so, swap it for the role, and carry on without it. Coaching a goal never requires knowing who someone is, and goals aimed at outcomes rarely name a person.
Zero data retention
A model has to read your goal to coach you on it. What it must not do is keep it. This product only uses models served under a zero data retention agreement: the provider processes your text to generate the reply and then permanently deletes it, and never trains on it. Three things are true of every AI call this site makes.
Nothing is kept
The provider deletes your prompt and the reply once the request completes. There is no transcript to subpoena, leak or hand to anyone later.
Nothing is trained on
Your goal never becomes training data. Tomorrow’s model will not have learned anything from your strategy.
No fallback to a provider that would
The gateway only routes to providers under a verified ZDR agreement. If none can serve the model, the request fails rather than quietly going elsewhere.
This isn’t a setting on a dashboard you have to take our word for. Every request the coach makes carries zeroDataRetention and disallowPromptTraining, in source anyone can read. Vercel’s AI Gateway treats those as filters rather than preferences: it routes only to providers it holds a verified ZDR agreement with, and if none can serve the model it refuses the request outright. So the failure mode is the coach falling back to the structural check and telling you why — never your words quietly going somewhere that would keep them.
What zero data retention doesn’t cover, so you know the edges: the web request that carries your draft to this site is still logged by the host, as any web request is; and a skill you download runs in your own AI assistant, under whatever retention terms you have with them, not ours.
The Outcome Coach can read a goal off a photo or a screenshot. That recognition runs inside your own browser — the picture is never uploaded and never reaches this site, so the faces, names and handwriting that end up in the corner of a whiteboard shot stay on your device. Only the words it finds go into the box, and only if you then submit them.
Holding us to it
“Refuses personal information” is one of the community principles, so it is reviewable like anything else here: this site is open source, and any feature that would collect personal data has to be argued for in public before it ships. If you think something on this site collects more than this page claims, raise it — that challenge is always welcome. Security contact and controls are in SECURITY.md.