The 114-entry directory is written for Chinese readers
Awesome Zhuiju Free collects 114 media resources under one Chinese README and a companion site. Its categories include 49 online-video sites, 16 magnet or BitTorrent indexes, 15 TVBox configuration sources, several search and subtitle tools, and 16 open-source projects. The repository has no English README in its 71-file tree. An English reader can translate the page, though labels such as access restrictions and copyright risk need more care than a browser's one-click translation gives them.
This is a directory rather than a media application. It sends readers to third-party sites and keeps the main records in resources/resources.json. The project says it does not host films, accounts, credentials, cracked software, or links to individual titles. That boundary makes the repository useful for discovery and easier to inspect, while leaving playback, registration, downloads, and any payment interaction on sites that the maintainer does not control.
A green mark means one homepage answered, not that a film played
The weekly release for 2026-W35 counted 71 accessible resources, 19 with restricted access, 8 unreachable, and 0 untested. A scheduled workflow updates those states each day around 09:00 Beijing time. New or changed records also trigger a check. This is more useful than an abandoned bookmark list because each row carries a date and the full result is preserved in reports/availability.json.
The README defines the limit clearly: the job checks whether a homepage responds from a GitHub Actions node. It does not watch a film, inspect every advertising redirect, test a download, or reproduce a reader's network. One row can be green in Actions and blocked by an ISP; another can open normally while exposing intrusive ads after the first click. Treat the status as a quick reachability signal and read the separate risk fields before visiting.
What happened when we ran it
Our sandbox installed commit 48e3cfb in 37 seconds, adding 35 packages and occupying 37 MB. The build completed in 8 seconds. Pip-audit found 0 known vulnerabilities in the installed Python environment. The 0.9 MB checkout contained about 2,024 lines of source across 71 files, which is small enough for a maintainer to review alongside the JSON data and generated README.
There was no test script or target, so our lab skipped tests. The repository also had no tests directory and no Dockerfile. It did have 5 CI workflow files covering availability, data validation, open-source statistics, issue conversion, and weekly releases. Those workflows are useful operational checks, but our supplied measurement does not report test cases or passed assertions. A downstream consumer should run the schema validator on every imported snapshot.
The JSON tracks 4 kinds of risk without deciding them for you
Each resource has a stable ID, category, URL, source, verification summary, and optional GitHub metadata. The schema separates capability scores from 4 risk areas: copyright, safety, privacy, and payment. A maintainer can sort or render the records without scraping the README table. Historical verification and notice files record changes and complaints, while a rights-holder issue template gives removal requests a defined path.
The ratings still involve judgment. The README scores breadth, speed, cleanliness, and stability, then marks each risk low, medium, high, or unknown. Issue 17 shows why the distinction matters: its submitter disclosed Google advertising, analytics, optional login, user-supplied M3U data, and uncertain channel rights. The directory can preserve that disclosure. It cannot audit every playlist or promise that a third-party operator will behave the same tomorrow.
Open submissions expose both useful detail and uneven evidence
GitHub listed 11 open issues and 2 pull requests on August 31, 2026. Most open issues are resource nominations. Some include a region, test date, ad behavior, and an explicit relationship disclosure. Others say little more than 'no risk.' That difference is visible instead of being hidden behind a staff rating, but readers must judge how much evidence supports each proposed entry. Inclusion should not turn a submitter's unchecked statement into a safety claim.
The contribution process helps. Templates ask for the URL, category, known risks, location, date, and relationship. Maintainers can validate the JSON before merging and retain a record of later complaints. CC BY 4.0 also permits reuse with attribution. For teams building another guide, the structured record and provenance fields are more useful than copying 114 rendered table rows.
Daily pushes show maintenance; three months do not prove durability
The repository was created on June 5, 2026 and last pushed on August 31, 2026. GitHub showed 7,388 stars, 418 forks, and 13 open issues and pull requests combined. The latest weekly release was published on August 30 and recorded additions, removals, adjustments, and availability changes. Current issue submissions show that readers are still feeding the list.
That activity earns the directory a place in a Chinese-speaking research workflow. Its age still matters because streaming domains change quickly and community ratings take time to earn trust. Use the JSON as a dated discovery dataset, preserve the verification timestamp, and follow local law plus each service's terms. The 5 workflows can tell you that a page answered; they cannot tell you whether using what lies behind it is a sound decision.