Mistake-Driven Remediation Generator
A fresh remediation set built from error mechanisms rather than copied wrong answers or generic extra questions.
Works on any free AI · Best tool: Any AI chat (ChatGPT, Claude, Gemini)
✅ How to use this effectively
- 1. Remove names and paste representative error work.
- 2. Check whether the diagnosis names a mechanism rather than a student trait.
- 3. Review the worked contrast and fresh items.
- 4. Reassess with a transfer item before closing the gap.
🔧 If it is not right, reply with this: If wrong, reply with: "Diagnose the first reasoning break, stop copying surface numbers and generate practice that isolates that mechanism."
THE PROMPT
ROLE: Act as an error-analysis specialist and targeted-practice designer.
CONTEXT: A teacher has anonymised errors from [SKILL] and needs fresh practice for the underlying misconceptions.
INPUTS — FILL EVERY PLACEHOLDER:
- [CLASS]
- [SKILL]
- [ANONYMISED ERROR WORK]
- [CORRECT METHOD]
- [ITEM COUNT]
- [REASSESSMENT DATE]
SOURCE-BOUND TOOL FACTS AND EXACT URL WHITELIST:
- [KHAN_ACADEMY] Whitelist: https://www.khanacademy.org/search?page_search_query={ENCODED_KEYWORDS}. Use search only; never invent opaque Khan topic or exercise paths.
- [WOLFRAMALPHA] Whitelist: https://www.wolframalpha.com/input?i={ENCODED_PLAIN_CALCULATOR_QUERY}. It recomputes a claim but does not prove surrounding reasoning; steps may require Pro.
DO THIS:
1. Locate the first reasoning break in each error sample without diagnosing the learner.
2. Group errors by mechanism such as sign, inverse operation, representation or domain.
3. Write one correct-versus-incorrect contrast for each mechanism.
4. Generate fresh minimal-pair items that change only the feature being relearned.
5. Add mixed transfer items only after the isolated checks.
6. Create a reassessment rule based on observable work, not score alone.
WORKED LINK EXAMPLE — COPY THIS EXACT ENCODING PATTERN:
Example task: Create remediation after students solve 3x+2=14 as x=16/3.
Expected result or check target: The practice isolates inverse operations before moving to mixed linear equations.
check this yourself
CHECK → https://www.wolframalpha.com/input?i=solve%203x%2B2%3D14
(paste-fallback: solve 3x+2=14)
Encoding note: The sample check decodes exactly and confirms x=4 without diagnosing why a learner erred.
OUTPUT FORMAT:
1. Error Mechanisms — first reasoning break by sample
2. Worked Contrasts — correct and incorrect paths
3. Targeted Practice — minimal pairs by mechanism
4. Transfer Check — mixed novel applications
5. Reassessment Rule — evidence for moving on
QUALITY AND TOOL GUARDRAILS:
- Never include student names, roll numbers or sensitive context.
- Do not label a learner by an error pattern or infer a diagnosis.
- Never invent Wolfram quiz slugs or Khan topic URLs; use whitelist rows or search links only.
- Label generated questions ORIGINAL PRACTICE.
- A correct answer alone does not show the misconception has been repaired; inspect the method.
- Copy every [PLACEHOLDER IN SQUARE BRACKETS] character-for-character in every template, code block and translated variant.
- Never output a desmos.com URL with parameters, a mathsolver.microsoft.com URL, an invented GeoGebra material ID, an invented DIKSHA code, an invented Khan topic path, or an unlisted quiz or simulation slug.
- Never describe an AI answer as verified. Every external check must be labelled “check this yourself” for the teacher or student to inspect personally.
- Do not invent exam facts, source citations, marks distributions, tool IDs or inaccessible links. Separate mathematical proof from computational checking.
=== TOOL-LINK OUTPUT CONTRACT — follow every rule exactly ===
1. You may output clickable links ONLY to: wolframalpha.com/input?i= · symbolab.com/solver?query= ·
geogebra.org/graphing|geometry|3d|cas?command= · math.stackexchange.com/search?q= · oeis.org/search?q= ·
artofproblemsolving.com/community/q1_ · khanacademy.org/search?page_search_query= ·
overleaf.com/docs?encoded_snip= · latex.codecogs.com/png.image? · docs.google.com/forms/...viewform?usp=pp_url
(Forms only with entry.N IDs I paste in) · plus any URL I give you or that appears in this prompt's whitelist table.
2. NEVER output: desmos.com links with query strings (they load blank), invented geogebra.org/m/ IDs, Khan topic
URLs, DIKSHA dial codes, Wolfram quiz/demo slugs not in my table, or mathsolver.microsoft.com (dead site).
3. Write queries in plain calculator syntax (x^2, sqrt(x), pi, 3/4) — never LaTeX — EXCEPT Overleaf and CodeCogs
payloads, which are LaTeX.
4. Percent-encode the query part: space→%20, +→%2B, ^→%5E, =→%3D, [→%5B, ]→%5D; in LaTeX payloads also
\→%5C, {→%7B, }→%7D, %→%25, #→%23, &→%26, newline→%0A. GeoGebra ?command= exception: encode ONLY + as %2B;
leave ( ) = ^ ; as typed.
5. One link per line, formatted exactly: CHECK → <full raw URL> (no markdown brackets, nothing after the URL).
6. Directly below every link, print: (paste-fallback: <the un-encoded query or commands>)
7. Before printing any link, silently decode your own URL and confirm it reproduces the query character-for-character;
rebuild it if not. If a URL would exceed 1,800 characters (Overleaf: 6,000), print only the paste-fallback.
8. Label every link "check this yourself" — never claim the link proves your answer is correct.
9. On phones these links must be opened in a browser, not an installed app — say this once at the end.