A catalog, not an application
BetterGI Scripts List is the content library behind a large part of the BetterGI ecosystem. Its four top-level collections contain combat strategies, JavaScript programs, map-following routes, and Genius Invokation card strategies. The generated repo.json index and linked bgi.sh site turn that material into a browsable catalog. Nothing in this repository runs by itself. BetterGI is the host that supplies screen capture, computer vision, simulated keyboard and mouse input, route execution, settings, and scheduling.
The distinction matters when judging setup and quality. Installing one script is not like installing a conventional package with a single supported API. A JavaScript entry can declare its main file, settings interface, saved data, author, version, and allowed HTTP destinations. It can call BetterGI functions for clicks, keys, screenshots, teleportation, routes, OCR, notifications, and other game actions. Combat and card scripts use their own domain formats, while pathing entries encode coordinates and movement modes.
This breadth is why the catalog is useful. Players can find targeted automations for daily commissions, fishing, cooking, gathering, artifact work, music, bosses, and many narrower jobs without waiting for every idea to enter BetterGI itself. Each entry has different authors, dependencies, test coverage, and assumptions.
BetterGI sets a demanding baseline
Before choosing a script, the user must have BetterGI working. Its main README requires 64-bit Windows 10 or newer and .NET 8. Computer vision is performance-sensitive, and the project recommends hardware that can keep Genshin running at 60 frames per second on medium settings. Only 16:9 displays are supported, with 1920x1080 windowed play recommended. HDR, driver filters, altered brightness, window changes, and monitor changes can break recognition or require a restart.
BetterGI also asks for administrator privileges because Genshin itself runs elevated and simulated input otherwise cannot reach it. Security software may block that input and may need an exception. A script that behaves correctly on the author's machine can miss a button after a game patch, language difference, DPI change, lag spike, or capture failure.
Most importantly, the host application's FAQ does not promise account safety. It says BetterGI avoids modifying game files or memory and works through vision plus simulated input, but also acknowledges that the game's terms name third-party software and simulated actions as possible ban reasons. Anyone who cannot accept that risk should stop before considering this catalog.
Curation is serious, but script quality still varies
The repository's submission guide is much more disciplined than a random scripts folder. Combat strategies need tested character naming and usage notes. Paths follow detailed material, region, ordering, and quantity conventions. JavaScript packages need manifests and readable source for HTML overlays. Resources must be explained, and promotional overlay content is prohibited. Internal review guidance tells reviewers to match declared network access, files, settings, minimum BetterGI versions, and high-risk behavior such as remote code or system operations.
There is useful mechanical validation too. The pathing validator checks JSON structure, coordinate fields, movement modes, supported actions, parameter formats, and minimum BetterGI versions. These checks improve consistency, but the project's own review notes correctly warn that parseable JSON does not prove a route behaves correctly in the game.
Current bug reports show the gap between structural validity and runtime safety. Issue 3530 identifies a deterministic JavaScript error in AutoPickLitter: an asynchronous refresh helper is used without await, causing repeated tasks, while local time replaces the Chinese server's UTC+8 boundary outside that timezone. Issue 3529 reports a cooking script continuing to click after ingredients are exhausted. Issue 3062 records a ley-line automation reaching oneDragonMode is not defined after several recognition and configuration errors. These are specific scripts, not evidence that every entry fails. They are evidence that supervision matters.
The contribution pipeline needs repair
One repository-level weakness is harder to isolate. The JSON validation workflow runs on pull_request_target, grants content and pull-request write permissions, checks out the contributor's head commit with allow-unsafe-pr-checkout, then executes build/validate.py from that checkout. A malicious pull request can modify that Python file while also touching a matching path JSON file. This is the exact class of untrusted-code execution that privileged pull-request workflows should avoid.
The repository's own AGENTS.md says workflows must not use write tokens or secrets to execute pull-request-controlled code, yet the current workflow does. Until maintainers change it to run trusted base-branch validation with minimal permissions, users should treat the catalog's distribution pipeline as carrying avoidable supply-chain risk. This does not prove the main branch has been compromised. It means the guardrail is weaker than the repository's review policy intends.
Fast-moving health with no stable release line
The last push was August 16, 2026. New scripts, route updates, fixes, bug reports, and reviews were arriving and closing throughout the same day. GitHub showed 34 open issues and 19 open pull requests, while the repository's numbering had passed 3,500.
There are no tagged GitHub releases. For a rolling content repository distributed through an index and mirror, that is not a sign of neglect. It does mean users lack a simple stable version marker for the whole collection. Script manifests carry individual versions, and compatibility can move with BetterGI and the game itself. Save a known-working script package and configuration before updating.
The absence of a repository license is a separate concern. Public source is visible, but visibility alone does not grant normal rights to redistribute or adapt every contribution. Fork maintainers, mirror operators, and developers borrowing code should obtain clarity first.
BetterGI Scripts List is worth using when one specific script solves a repetitive task that BetterGI does not already handle. Read that script's Chinese documentation, inspect its manifest and source, confirm the minimum host version, and watch its first runs. Avoid installing the catalog mentally as one trusted product. Its strength is community variety, and that same variety is the reason every script earns trust separately.