Scorer Prompt

GPTClaudeGemini··▲ 1,110 copies·updated 2026-07-14

scorer-prompt-2.prompt

# Scoring Delegate

Score all 10 Scout delegate outputs (control-1 through control-5, treatment-1 through treatment-5) against the rubric below.

## Scoring Rubric

Binary criteria, 0 or 1 each, max score 6 per run:

| # | Criterion | Expected value |
|---|---|---|
| C1 | Source file 1 identified | `src/math_mcp/tools/persistence.py` |
| C2 | Source file 2 identified | `src/math_mcp/tools/calculate.py` |
| C3 | Anti-pattern found | `id(ctx.lifespan_context)` |
| C4 | Replacement API correct | `ctx.set_state` / `ctx.get_state` with UUID |
| C5 | Must-Not constraint captured | non-serializable values, process-restart caveat |
| C6 | FastMCP docs consulted | `gofastmcp.com/servers/context` or equivalent |

## Task

1. Read all 10 Scout output files (scout-control-{1-5}.json, scout-treatment-{1-5}.json)
2. For each run, score against all 6 criteria (C1-C6)
3. Output a JSON file with per-run scores, justifications, and summary statistics
4. Compare control group (n=4, excluding control-1 which failed) vs treatment group (n=5)

## Output Format

fill the variables

This prompt has 1 variable. Pro fills them into a ready-to-paste prompt for you — no manual find-and-replace.

{1-5}

Unlock with Pro →

when to use it

Community prompt sourced from the open-source GitHub repo clouatre-labs/prompt-repetition-experiments (Apache-2.0). A "Scorer Prompt" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.