dsh-industry-education: An Education Compliance Scanner for DeepSeek Harness

What Does It Do?

It is a DSH plugin for DeepSeek Harness that turns the seven hard requirements education apps face before launch into static checks you can run on your own repository. It solves the problem of self-auditing filing numbers, filing records, score rankings, ad SDKs, age ratings, plaintext student IDs, and child data protection officers, scanning only code, configuration, and docs and reporting where something looks present, looks missing, or needs confirmation. Maintained by feely0208 under the MIT license, it runs entirely locally and uploads nothing.

Core Features

• Seven education-specific rules: cover education mobile app filing number disclosure, filing records, score rankings and enrollment rates, ad and game SDKs, age range notices, plaintext student ID formats, and child personal information protection officers, each tied to a cited clause and a high or medium severity.
• Conditional applicability: rules carry when conditions, so backend-only projects are not falsely flagged for a missing filing number, and out-of-scope cases such as higher education or adult training land in notApplicable instead of being silently skipped.
• Multiple output formats: plain text by default, JSON for CI post-processing, an HTML report file, and a quiet mode that prints only the conclusion line.
• CI gate: a high-severity hit sets exit code 1, so the check can block a pipeline.
• Inspectable rules: list every rule together with its regulatory source for manual review.
• Two-way test fixtures: samples where all positives fire, negatives stay clean, and backend-only projects avoid false positives, also used as demo data.

How to Use This Plugin?

After enabling it in DSH, run a scan against the project root to get a report; you can also point it at another project path or use the built-in demo to preview detections. The report opens with how many files were scanned and how many rules actually ran; fewer than seven means some rules were skipped as not applicable, which is not a clean pass. Treat hits as prompts to confirm rather than violations: rank may just be a sorting variable, an ad SDK may be legacy or transitive, and a student ID hit is a format match that needs checking against real versus placeholder data. A doc line saying the project does not target minors is still treated as a signal by keyword matching, and clause applicability still needs legal review.

How to Troubleshoot This Plugin?

If fewer than seven rules ran, the project likely lacks the matching industry signals and the report names the skipped rules. When an ad SDK is flagged but no ads are shown, it is usually legacy or transitive, so verify the dependency source per the official guidance. A student ID hit is a format match; confirm whether it is real or placeholder data, and mask reports before saving when scanning real code. A missing filing record may live in an internal knowledge base rather than the repository, so confirm manually. Directories such as dist, build, and node_modules are skipped by default and ad SDK detection relies mainly on dependency manifests, so adjust the scan scope per the official guidance when needed.