FICTIONAL SCHOLARSHIP REVIEW PRACTICE
Every application, rule, scoring guide, and sample answer here is made up for practice.
A17, A18, and A19 are made-up applicant codes used instead of names. Each code refers to the same applicant across the files. For example, A17-transcript.pdf is the pretend transcript filename for applicant A17.
A manifest is a file list. This list names pretend PDF files; those PDFs are not included. A listed name does not tell us what is inside a document. In real applications, replacing names with codes does not remove every detail that could identify someone.

Words used in the prompts: a rubric is a scoring guide. A criterion is one part being judged. An anchor describes what earns a score. A provisional score is a suggestion for a person to check. A source pointer tells you where to find the original information.

1. Open ChatGPT. Create a practice Project if available, or start a fresh chat.
2. Add project-instructions.txt as Project instructions, or paste it into the chat.
3. Upload only the inputs named in each prompt. The prompt files give the source boundaries.
4. Run one demonstration at a time. Compare the response with prepared-examples.md and the source files.
5. Record errors and corrections in review-log.csv. Staff retain every consequential decision.

Prompt 1: manifest.csv and applications.csv.
Prompt 2: program-rules.md, manifest.csv, applications.csv.
Prompt 3: statements.md and rubric.md.

Source pointers: CSV data row numbers below exclude the header. Manifest A17 rows 1-4, A18 rows 5-7, A19 rows 8-12. Applications A17 row 1, A18 row 2, A19 row 3.

No paid API or special model is required by this website. ChatGPT feature availability and account limits vary. No real applicant data belongs in this public exercise.
