The project is stable: no new features, code changes only for security advisories and runtime/dependency updates, and issues and pull requests are for the data (add, correct, remove a game). - README: status, build, quality, version, licence and runtime badges, a "Project status" section, and a data-first Contribute section - CONTRIBUTING, SECURITY, Contact, the PR template and CLAUDE.md state the same policy - Issue forms: add_game and edit_game are new; feature_request, feedback and off_topic are removed; blank issues are off (config.yml) - About page: "Report something" links the five remaining forms Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
4.8 KiB
Security policy
What this project actually is
The previous version of this file said the repository was "just a Markdown list… no executable code, no dependencies, no backend, no user data." That was wrong on every count, so here is the real picture:
- a Cloudflare Worker serving the site. It holds a D1 binding, an R2
bucket for the game art (public at
media.free-steam-games.win), a 15-minute cron (reconcile, auto-approve, and minting the art into that bucket), and, for the admin path, a credential that can write to this repository - a SvelteKit / Svelte 5 web app, prerendered to static pages, with a few
dozen npm dependencies, plus the
/adminapp embedded in the Worker - a Tauri 2 desktop and Android app, signed, with an auto-updater
- a Python pipeline run by GitHub Actions. The data jobs push to
mainwith the built-inGITHUB_TOKEN; the release jobs hold the app signing keys
Operational hardening (headers, HSTS, rate limiting, WAF) is documented in
docs/SECURITY_SETUP.md, with a check command for
each item.
There is genuinely something to attack here. Hence a real policy.
Reporting
Use GitHub's private vulnerability reporting, which is enabled on this repository: Security → Report a vulnerability. It is private between you and the maintainer.
If that does not work for you, email security@poli0981.dev. The same
address is published in AUTHORS.md,
docs/Contact.md, username.txt and
https://free-steam-games.win/.well-known/security.txt.
That security.txt is the machine-readable version of this policy
(RFC 9116). Its Policy field points back at this file and its Canonical
field at itself; if the two ever disagree, this file wins and the
security.txt is stale.
Please do not open a public issue for anything exploitable. Public issues are fine for a dead link or a game that is no longer free — those are data problems, not security ones.
Useful report: what you did, what happened, why it matters, and how to reproduce. A proof-of-concept URL beats a paragraph of description.
In scope
- The Cloudflare Worker (
web/worker/**) — the data proxy, the image proxy, and anything reachable under/api/* - The
/adminpath and/api/ingest/*, and their Cloudflare Access gates - The D1 database and anything that writes to it
- The media host,
media.free-steam-games.win(the R2 bucket's public domain), and whatever the cron writes to it - The GitHub Actions workflows, especially any path where untrusted input
reaches a step that can push to
main - The Tauri updater, its signing key, and its feed at
/api/updates/desktop - Dependency vulnerabilities that are actually reachable from shipped code
Things worth probing specifically: making /img/* fetch an origin that is not
shared.akamai.steamstatic.com or shared.fastly.steamstatic.com (or, for
the avatars under /img/gh/*, avatars.githubusercontent.com); making
/api/data/* return a path outside its allowlist; reaching admin
functionality without an Access session; getting the update feed to offer a
file that is not this repository's release download.
Out of scope
- Games in the list being bad, predatory, or full of microtransactions. Take it up with the developer.
- Anything about Valve or Steam itself.
- Missing security headers with no demonstrated impact, and scanner output pasted without a working exploit.
- Denial of service by volume.
- Getting past the website's Turnstile human check from your own browser. It is
a client-side gate on purpose: the pages are static and the catalogue API is
public by design (
docs/ToS.md§9), so it guards nothing a script cannot already read. A way to make/api/human-checkaccept a token Cloudflare rejected is in scope. - Social engineering of the maintainer.
- Accessibility barriers. They are real bugs, but not security ones: see
ACCESSIBILITY.md.
What to expect
One person, as a hobby, in a timezone that is probably not yours. A first reply usually within a week. Serious issues get fixed before they are discussed publicly; minor ones may sit for a while.
No bounty programme, no money — the disclaimers about the maintainer being
broke are not a joke. Credit in docs/ACKNOWLEDGEMENTs.md if you want it.
Supported versions
The deployed site and main are the only supported versions. Desktop and
Android releases are supported at the latest tag only; there are no backports.
The project is stable and in maintenance mode (see
CONTRIBUTING.md): no new features are developed, and
security fixes and runtime / dependency updates are exactly what it is still
maintained for. Dependabot, CodeQL and SonarCloud keep running, and a fix
ships as a new site deploy and, when the apps are affected, a new app release.