A folder of documents is not a prepared submission.
Payslip-heavy preparation involves more than collecting PDFs. Documents can overlap, appear to leave gaps, contain sensitive details or belong to different people. Automating every decision risks turning an extraction error into a misleading final pack.
SkillAssist approaches this as a guided document-preparation workflow for a TRA-oriented use case. Its product boundary is preparation and review; the implementation does not promise an assessment outcome.
Make the user’s review a first-class step.
The workflow moves from account and profile onboarding to occupation selection, payslip workspaces, upload, extraction, review and final export. Users retain control over document inclusion and labels.
Processing status and review status are separate. A file can finish extraction and still need a person’s attention. This distinction prevents technical completion from being mistaken for document readiness.
Build the foundations around document ownership.
The Next.js and PostgreSQL foundation includes authentication, protected workspaces, validated upload intake and processing orchestration. Text-based PDFs can produce structured extracted fields and warnings; image uploads remain reviewable with explicit extraction limitations.
Workspace analysis records likely duplicates, timeline gaps and possible multi-person sets as review findings. PDF selection and protected combined exports extend the workflow. Later foundations add individual credit purchases through verified Stripe webhooks, professional-help requests, forms and product support.
Keep uncertain automation explainable.
Analysis does not automatically delete, exclude or split documents. Sensitive-data detection stores logical redaction targets for review, rather than claiming that final PDF masking has happened. The current combined export does not apply rendered redaction.
The billing foundation grants credits after verified payment and consumes an export credit only after a successful completed pack. Failed exports and repeat downloads do not carry the same charge. These choices align the workflow with a completed user action rather than an attempted processing step.
Production readiness is its own milestone.
The repository describes a foundation, not a fully validated production service. Local file storage, limited text-PDF extraction, reviewer allowlists and PDF-only combined exports define important current boundaries.
Proposed next steps include production storage, broader extraction evaluation, rendered redaction with output verification and fuller administrative roles. Growth should follow evidence that ownership, review and export behaviour remain dependable across varied document sets.
What exists. What comes next.
A substantial document-preparation foundation, with explicit review stages and ownership checks. Production storage, richer extraction and rendered redaction remain future work.
This case study draws on the project repository and available build history. Problem framing and proposed improvements are editorial synthesis; unverified results are not presented as measured outcomes.
Some repositories are private; access is required to view those references.