← ClaudeAtlas

prompt-coachlisted

Use when the user asks "rate my prompt" / "evaluate my prompt(s)" / "how am I prompting?" / "give me prompt feedback" / "score my last prompt" / "4Ds analysis" / "prompt retrospective". Also use when the user asks for an end-of-session review of their work-with-Claude habits, OR after a session involving 20+ turns where the user explicitly wants improvement signal. NEVER fire unsolicited — prompt critique without consent is unwelcome.
damson/hard-won-skills · ★ 1 · AI & Automation · score 67
Install: claude install-skill damson/hard-won-skills
# Prompt Coach — score prompts on the 4Ds framework This skill scores the user's prompts on four dimensions, identifies the weakest, and proposes a rewrite. The goal is to make the user a more efficient operator over time — not to lecture. Sibling skills: - `skill-opportunity-finder` — surfaces repetition patterns to skillify. - `prompt-coach` (this one) — retrospective scoring with concrete rewrites. ## The 4Ds framework | Letter | What it measures | |---|---| | **D**elegation | Trust in the assistant to choose how, not just what. | | **D**escription | Clarity of subject + scope + expected output. | | **D**iscernment | Quality of feedback on the assistant's output — pushing back, validating, catching mistakes. | | **D**iligence | Follow-through across turns — reminders, verification, applying new conventions retroactively. | Each scored **1–5** against the anchors table below. ## Procedure ### Mode A — single prompt When the user pastes / quotes one prompt and asks "rate this": 1. Read the prompt + the ~3 turns of session context around it. 2. Score each D from 1 to 5. Justify each with the specific cue in the prompt (or its absence). Diligence is cross-turn by definition: in Mode A score it only when the prompt itself follows up on an earlier commitment; otherwise N/A. 3. Identify the **weakest D** and explain it in one sentence. 4. Propose **one rewrite** of the prompt that fixes that weakest dimension without inflating word count by more than ~30% — the one lengt