evaluator-be-corrective

name	evaluator-be-corrective
description	Grade a backend corrective (debugging) benchmark submission under submissions/be-corrective/ by running each question's Node check.mjs (binary 10/0). Use when asked to evaluate, score, or grade a be-corrective submission.
disable-model-invocation	true

Evaluator (BE Corrective)

Grade one be-corrective submission in submissions/be-corrective/<name>/. You (and only you) may read reference-answers/be-corrective/. Per-question grading is fully scripted — run the checkers and record output; no LLM matching. No Docker, no external DB.

Scoring

• Each of the nine questions: 10 if check.mjs exits 0 and prints PASS; 0 if FAIL or missing deliverable.
• Overall: sum across nine (max 90).

Inputs

Question	Submission	Checker
q1-api-easy/mid/hard	submissions/be-corrective/<name>/q1-api-*/	reference-answers/be-corrective/q1-api-*/check.mjs
q2-db-easy/mid/hard	submissions/be-corrective/<name>/q2-db-*/	reference-answers/be-corrective/q2-db-*/check.mjs
q3-service-easy/mid/hard	submissions/be-corrective/<name>/q3-service-*/	reference-answers/be-corrective/q3-service-*/check.mjs

Pass the submission directory (absolute path) as the sole argument.

Run checkers (required)

node reference-answers/be-corrective/q1-api-easy/check.mjs  <abs>/submissions/be-corrective/<name>/q1-api-easy

(Repeat for all nine ids.) Exit 0 + PASS → 10; non-zero + FAIL → 0. Copy the first failure reason into the note. Node stdlib only is required for the submission; checkers need no extra install beyond the repo’s Node.

Output

Write into submissions/be-corrective/<name>/:

EVALUATION.md

Per-question table: question | kind=run | score(0/10) | check output. End with overall (sum) and a short summary.

scores.json

{
  "submission": "<name>",
  "set": "be-corrective",
  "questions": {
    "q1-api-easy":     { "kind": "run", "score": 0, "note": "" },
    "q1-api-mid":      { "kind": "run", "score": 0, "note": "" },
    "q1-api-hard":     { "kind": "run", "score": 0, "note": "" },
                 
                  
                 
            
             
            
  
   

Report the overall score and which questions still failed.