Skip to main content

prompt-check

Evaluates the quality of the user's prompts to Claude. Auto-triggers when the user appears to be prompting poorly — vague questions, shotgun debugging, desperation signals like rapid rephrasing, 'why isn't this working', 'I've tried everything', 'please just do X', or questions that lack specificity about what they're actually looking for. Also triggers on direct requests like 'check my prompt', 'am I asking this right', 'how should I prompt this'.

Quellinformationen

Repository
devinat1/learning-toolkit
Letzte Quellaktivität
11. April 2026 um 22:26
Erkannte Sprache von SKILL.md
Englisch
Sterne
1
Forks
0

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

SKILL.md wird angezeigt

SKILL.md
Quellanweisungen · Schreibgeschützte Vorschau
name
prompt-check
description
Evaluates the quality of the user's prompts to Claude. Auto-triggers when the user appears to be prompting poorly — vague questions, shotgun debugging, desperation signals like rapid rephrasing, 'why isn't this working', 'I've tried everything', 'please just do X', or questions that lack specificity about what they're actually looking for. Also triggers on direct requests like 'check my prompt', 'am I asking this right', 'how should I prompt this'.
# Prompt Check Evaluates prompt quality, diagnoses problems, suggests rewrites, and coaches toward better prompting habits. ## Phase 1 — Desperation Rating Evaluate the user's prompt and assign a desperation level 1-10: | Level | Signal | Meaning | |-------|--------|---------| | 1-3 | Clear, specific, well-scoped | You're in control — minor suggestions at most | | 4-6 | Somewhat vague, missing context, could be interpreted multiple ways | You're drifting — let's sharpen this | | 7-9 | Throwing things at the wall, rapid rephrasing, "just make it work" energy | You're at the mercy of the LLM — stop and regroup | | 10 | Copy-pasting error messages with no context, "please help" | Full desperation — we need to back up significantly | Present the rating with a brief explanation of what signals were detected. ### Low Desperation (1-3) — Light Touch At this level the skill should be near-invisible. Do not interrupt the user's flow with a full analysis. A one-liner like "Good prompt — clear and specific" followed by actually answering the question is the right weight. The full phase breakdown (diagnosis, rewrite, coaching) only activates at level 4+. ## Phase 2 — Problem Diagnosis Only activates at desperation level 4+. Identify what's specifically wrong with the prompt: - **Missing concepts** — the user is asking about X but doesn't seem to understand prerequisite Y - **Lack of specificity** — the question could mean 5 different things, and the LLM will guess - **Wrong framing** — the user is asking how to do Z but Z isn't actually what they need - **Context gaps** — the prompt assumes knowledge that wasn't provided Each issue is stated concretely, tied to the actual prompt. No generic advice. ## Phase 3 — Rewrite Suggestion Only activates at desperation level 4+. Offer a rewritten version of the prompt that would get better results. Show before and after so the user sees what changed and why. Format: > **Your prompt:** [original] > > **Sharper version:** [rewrite] > > **What changed:** [1-2 sentences explaining the improvement] This teaches better prompting. Do not answer the user's original question — just improve how they asked it. ## Phase 4 — Coaching **High desperation (7+):** - Help the user figure out what they're actually trying to accomplish - "I think you're actually trying to accomplish X — is that right?" - If the real problem is a knowledge gap, suggest switching to `learning-toolkit:learn` - Stay with the user until they've reframed their question **Moderate desperation (4-6):** - A sentence or two of guidance on what was missing - Let the user rephrase and try again - If they rephrase well, acknowledge the improvement and answer the question **Low desperation (1-3):** - Already handled in Phase 1 — no coaching needed
Auf GitHub ansehen