Prompt Quality Analyzer: AI Chat Log Assessment

chat_log: 'User: write me something about climate change. AI: [500 word essay]. User: no make it shorter. AI: [300 word summary]. User: actually I need it for a tweet. AI: [280 char tweet]. User: can you add some stats. AI: [tweet with stats added]' focus_area: instruction clarity

## Prompt Quality Assessment Report ### Executive Summary This 4-turn conversation scored 3.8/10 overall. The core failure is absent upfront context: the user's true goal (a stat-backed tweet) was only revealed in turn 4, forcing three full AI regenerations. Every earlier response was wasted compute. With a single well-structured initial prompt, this conversation collapses to one turn. ### Intent Map | Turn | Stated Request | Inferred Intent | Gap? | |------|---------------|-----------------|------| | 1 | 'Write something about climate change' | Short, punchy, stat-driven tweet | HIGH | | 2 | 'Make it shorter' | Reduce to tweet length | MEDIUM | | 3 | 'For a tweet' | 280-char constraint | LOW | | 4 | 'Add stats' | Credibility/shareability | LOW | ### Conversation-Level Pattern Analysis - **Progressive Disclosure** (HIGH severity): User revealed constraints one per turn rather than upfront. Each new constraint invalidated prior output. - **Format Omission** (HIGH): No platform, length, or tone specified in turn 1. - **Missing Audience Context** (MEDIUM): No indication of who would read this or what action it should drive. ### Overall Scores | Dimension | Score | |-----------|-------| | Intent Clarity | 2/10 | | Context Richness | 3/10 | | Constraint Precision | 2/10 | | Outcome Effectiveness | 6/10 | | **Overall** | **3.8/10** | ### Rewrite Clinic **Original:** 'write me something about climate change' **Diagnosis:** Zero constraints, zero context, maximum ambiguity. **Rewritten:** 'Write a single tweet (under 260 chars) about climate change for a general audience. Include one striking statistic. Tone: urgent but not alarmist. Goal: drive retweets.' **Why it's better:** Specifies format, length, audience, tone, stat requirement, and success metric in one pass — eliminating all 3 follow-up turns. ### Skill-Building Takeaways 1. State your output format and platform in turn 1 — never assume the AI knows where content will live. 2. Collect all constraints before prompting; write them down first, then compose the prompt. 3. If you find yourself saying 'no, actually' in a follow-up, that constraint belonged in the original prompt.