- Home
- Skills
- Education & Learning
- Gradebook Gate: Catch Wrong Final Grades Before Release
Gradebook Gate: Catch Wrong Final Grades Before Release
**What you get.** A report with a verdict (READY / REVIEW / FIX BEFORE RELEASE) and one row per problem: student, column, the value found, the value to put back, and why.
$9
Works with the AI Agents you already use
Gradebook Gate: Catch Wrong Final Grades Before Release
Example session with this skill installed
Check my gradebook before I release grades. Homework 20%, quizzes 30% with the lowest quiz dropped, midterm 20%, final exam 30%; A is 90, B 80, C 70, D 60. File: gradebook-sample.csv (13 students, Final % and Letter columns).
- Read your context and instructions
- Compiled the gradebook gate
- Generated the data export
Gradebook Gate report
Verdict: FIX BEFORE RELEASE - 4 critical, 6 high, 0 medium, 1 info
- File: gradebook-sample.csv + policy-sample.json
- Students: 13 | score columns: 8 | recorded total columns: Final % | letter column: Letter
- Model: weighted (Homework 20%, Quizzes 30%, Midterm 20%, Final Exam 30%)
- Letter scale (policy): A >= 90, B >= 80, C >= 70, D >= 60, F >= 0
- 'Final %' read as percent, 1 decimal place(s), rule: blank/absent = 0; weights rescaled over graded categories; drop-lowest applied; 8 of 13 rows match the recomputation
Findings
| # | Severity | Student | Column | Found | Expected | Why | Fix |
|---|---|---|---|---|---|---|---|
| 1 | CRITICAL | Sara Mitchell (S-103) | Final % | 93.5 | 87.1% | Recorded value is reproduced exactly if 'Final Exam' is left out of both the marks and the maximum. On the scale the recorded total gives A, the corrected one B. | Replace with the expected value, then fix the formula for this row. |
| 2 | CRITICAL | Sara Mitchell (S-103) | Letter | A | B | Letter does not match the corrected Final % = 87.1% on the stated scale. | Correct the letter after fixing the total. |
| 3 | CRITICAL | Youssef Adel (S-104) | Final % | 71.3 | 74.1% | Recorded value is reproduced exactly if drop-lowest is not applied. | Replace with the expected value, then fix the formula for this row. |
| 4 | CRITICAL | James Carter (S-106) | Letter | C | B | Letter does not match Final % = 81.7% on the stated scale. | Correct the letter (or the policy scale, if the scale is wrong). |
| 5 | HIGH | (whole sheet) | ID | rows 4, 11 | one row per student | Same student id appears on 2 rows; one result may overwrite or double the other. | Merge or remove the duplicate row. |
| 6 | HIGH | Omar Farouk (S-102) | HW2 | 12 | <= 10 | Score is above the column maximum (typo, wrong column, or undeclared extra credit). | Fix the mark, or list the column under extra_credit_columns / set allow_over_max. |
| 7 | HIGH | Mariam Nabil (S-105) | Final % | 76.8 | 76.9% | Exact .5 was rounded half-to-even (banker's rounding, e.g. Python round() or ROUND_HALF_EVEN); school rounding is usually half-up. | Round half-up as the grading policy states; avoid half-to-even routines. |
| 8 | HIGH | نور الهدى سالم (S-107) | Final % | 84.4 | 83.6% | Recorded value is reproduced exactly if blanks/absences are left out instead of counted as 0; or if 'Quiz 1' is left out of both the marks and the maximum. | Replace with the expected value, then fix the formula for this row. |
| ... | (3 more findings, then boundary watch: Daniel Ruiz 89.6% - 0.4 below A) |
gradebook-sample.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
A gradebook can look finished and still be wrong for three students. The SUM range stops one column short on rows added late, drop-lowest was applied by hand to half the class, a blank counted as 0 for one student and was ignored for another, or a script's round() turned 84.5 into 84. Nobody re-adds 30 rows at the end of term, so the error surfaces as a grade appeal. LMS help forums (Canvas, Brightspace, PowerSchool) carry many "final grade doesn't match the items" threads.
What you get. A report with a verdict (READY / REVIEW / FIX BEFORE RELEASE) and one row per problem: student, column, the value found, the value to put back, and why. For every wrong total it names the cause that reproduces the recorded number exactly: a column missing from the sum, drop-lowest not applied, blanks or absences treated the other way, an excused mark counted as 0, weights ignored, banker's rounding or truncation. When a fix crosses a letter cut-off it says so, and letters are judged against the corrected total, so it never tells you to change a correct letter.
How it works. A standard-library Python script recomputes every student with exact fractions (no floating-point drift at 89.5):
- models: total points, per-column weights (WEIGHT row), or category weights from a small JSON policy, with drop-lowest and extra credit
- reads CSV, TSV and XLSX (saved values), including Canvas exports: the Points Possible row, Current vs Final Score semantics, each letter column paired with its own score column
- English and Arabic headers, Arabic-Indic digits, ٫ and decimal commas, marks such as EX, ABS, غ, غائب, معذور
- letter scales with any labels (A-F, plus/minus, ممتاز / جيد جدا ...), plus a scale-free check that catches two students on the same % with different letters
- catches weights that don't add to 100, scores above the maximum, duplicate students, unreadable cells
Checked in 3 rounds of independent QA, including 1,650 randomly generated students: zero false alarms on the correct gradebooks and every planted error caught (stale columns, skipped drops, rounding, wrong letters, excused-as-zero).
Limits. Values only: it does not read spreadsheet formulas, and XLSX cells need saved values. Expectations come from your policy, the sheet's MAX/WEIGHT rows, or the rule most rows follow; a wrong policy gives wrong expectations. Exports with no max row (Moodle's) need maximums in the policy. No curving or standards-based grading.
Reads values, not spreadsheet formulas; XLSX cells must have saved values. Expectations come from your policy, the sheet's MAX/WEIGHT/CATEGORY rows, or the rule most rows follow, so a wrong policy gives wrong expectations. Exports without a max row (e.g. Moodle) need maximums in the policy. Extra credit only where declared. No curving, standards-based or multi-term calculations. One sheet per run. Old .xls is not read.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed today
Requires
Permissions
File Scopes
Runs fully offline. No network calls, no subprocesses. Never modifies your gradebook. The XML namespace strings in the XLSX reader are identifiers, not network addresses.