| 100.00% | Adverbs in dialogue tags | Target: ≤10% dialogue tags with adverbs | | totalTags | 3 | | adverbTagCount | 0 | | adverbTags | (empty) | | dialogueSentences | 6 | | tagDensity | 0.5 | | leniency | 1 | | rawRatio | 0 | | effectiveRatio | 0 | |
| 100.00% | AI-ism adverb frequency | Target: <2% AI-ism adverbs (58 tracked) | | wordCount | 1042 | | totalAiIsmAdverbs | 0 | | found | (empty) | | highlights | (empty) | |
| 100.00% | AI-ism character names | Target: 0 AI-default names (17 tracked, −20% each) | | codexExemptions | (empty) | | found | (empty) | |
| 100.00% | AI-ism location names | Target: 0 AI-default location names (33 tracked, −20% each) | | codexExemptions | (empty) | | found | (empty) | |
| 76.01% | AI-ism word frequency | Target: <2% AI-ism words (290 tracked) | | wordCount | 1042 | | totalAiIsms | 5 | | found | | | highlights | | 0 | "footsteps" | | 1 | "electric" | | 2 | "constructed" | | 3 | "scanned" |
| |
| 100.00% | Cliché density | Target: ≤1 cliche(s) per 800-word window | | totalCliches | 0 | | maxInWindow | 0 | | found | (empty) | | highlights | (empty) | |
| 100.00% | Emotion telling (show vs. tell) | Target: ≤3% sentences with emotion telling | | emotionTells | 0 | | narrationSentences | 84 | | matches | (empty) | |
| 91.84% | Filter word density | Target: ≤3% sentences with filter/hedge words | | filterCount | 2 | | hedgeCount | 1 | | narrationSentences | 84 | | filterMatches | | | hedgeMatches | | |
| 100.00% | Gibberish response detection | Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words) | | analyzedSentences | 87 | | gibberishSentences | 0 | | adjustedGibberishSentences | 0 | | longSentenceCount | 0 | | runOnParagraphCount | 0 | | giantParagraphCount | 0 | | wordSaladCount | 0 | | repetitionLoopCount | 0 | | controlTokenCount | 0 | | maxSentenceWordsSeen | 39 | | ratio | 0 | | matches | (empty) | |
| 100.00% | Markdown formatting overuse | Target: ≤5% words in markdown formatting | | markdownSpans | 0 | | markdownWords | 0 | | totalWords | 1040 | | ratio | 0 | | matches | (empty) | |
| 100.00% | Missing dialogue indicators (quotation marks) | Target: ≤10% speech attributions without quotation marks | | totalAttributions | 2 | | unquotedAttributions | 0 | | matches | (empty) | |
| 100.00% | Name drop frequency | Target: ≤1.0 per-name mentions per 100 words | | totalMentions | 23 | | wordCount | 1023 | | uniqueNames | 10 | | maxNameDensity | 0.78 | | worstName | "Quinn" | | maxWindowNameDensity | 1.5 | | worstWindowName | "Quinn" | | discoveredNames | | Thames | 1 | | Raven | 1 | | Nest | 1 | | Herrera | 6 | | Quinn | 8 | | Soho | 1 | | Saint | 1 | | Christopher | 1 | | Tube | 1 | | Glock | 2 |
| | persons | | 0 | "Raven" | | 1 | "Herrera" | | 2 | "Quinn" | | 3 | "Saint" | | 4 | "Christopher" |
| | places | | | globalScore | 1 | | windowScore | 1 | |
| 63.79% | Narrator intent-glossing | Target: ≤2% narration sentences with intent-glossing patterns | | analyzedSentences | 58 | | glossingSentenceCount | 2 | | matches | | 0 | "quite running but wanted to be" | | 1 | "not quite people — selling from arrangements that Quinn's brain kept trying to rationalise and kept failing to" |
| |
| 100.00% | "Not X but Y" pattern overuse | Target: ≤1 "not X but Y" per 1000 words | | totalMatches | 0 | | per1kWords | 0 | | wordCount | 1040 | | matches | (empty) | |
| 100.00% | Overuse of "that" (subordinate clause padding) | Target: ≤2% sentences with "that" clauses | | thatCount | 1 | | totalSentences | 87 | | matches | | 0 | "understood that drawing" |
| |
| 100.00% | Paragraph length variance | Target: CV ≥0.5 for paragraph word counts | | totalParagraphs | 29 | | mean | 35.86 | | std | 31.83 | | cv | 0.887 | | sampleLengths | | 0 | 27 | | 1 | 79 | | 2 | 83 | | 3 | 9 | | 4 | 64 | | 5 | 3 | | 6 | 7 | | 7 | 2 | | 8 | 71 | | 9 | 15 | | 10 | 75 | | 11 | 12 | | 12 | 23 | | 13 | 45 | | 14 | 13 | | 15 | 93 | | 16 | 3 | | 17 | 22 | | 18 | 4 | | 19 | 7 | | 20 | 124 | | 21 | 58 | | 22 | 30 | | 23 | 48 | | 24 | 28 | | 25 | 44 | | 26 | 22 | | 27 | 10 | | 28 | 19 |
| |
| 92.73% | Passive voice overuse | Target: ≤2% passive sentences | | passiveCount | 3 | | totalSentences | 84 | | matches | | 0 | "was plastered" | | 1 | "been boarded" | | 2 | "was crossed" |
| |
| 100.00% | Past progressive (was/were + -ing) overuse | Target: ≤2% past progressive verbs | | pastProgressiveCount | 2 | | totalVerbs | 180 | | matches | | 0 | "were watching" | | 1 | "was doing" |
| |
| 100.00% | Em-dash & semicolon overuse | Target: ≤2% sentences with em-dashes/semicolons | | emDashCount | 2 | | semicolonCount | 0 | | flaggedSentences | 1 | | totalSentences | 87 | | ratio | 0.011 | | matches | | 0 | "Behind them stood people — and things that were not quite people — selling from arrangements that Quinn's brain kept trying to rationalise and kept failing to." |
| |
| 100.00% | Purple prose (modifier overload) | Target: <4% adverbs, <2% -ly adverbs, no adj stacking | | wordCount | 1025 | | adjectiveStacks | 0 | | stackExamples | (empty) | | adverbCount | 27 | | adverbRatio | 0.026341463414634145 | | lyAdverbCount | 6 | | lyAdverbRatio | 0.005853658536585366 | |
| 100.00% | Repeated phrase echo | Target: ≤20% sentences with echoes (window: 2) | | totalSentences | 87 | | echoCount | 0 | | echoWords | (empty) | |
| 100.00% | Sentence length variance | Target: CV ≥0.4 for sentence word counts | | totalSentences | 87 | | mean | 11.95 | | std | 9.06 | | cv | 0.758 | | sampleLengths | | 0 | 17 | | 1 | 10 | | 2 | 29 | | 3 | 8 | | 4 | 3 | | 5 | 39 | | 6 | 2 | | 7 | 1 | | 8 | 2 | | 9 | 23 | | 10 | 10 | | 11 | 3 | | 12 | 8 | | 13 | 34 | | 14 | 9 | | 15 | 26 | | 16 | 6 | | 17 | 5 | | 18 | 27 | | 19 | 3 | | 20 | 7 | | 21 | 2 | | 22 | 4 | | 23 | 31 | | 24 | 11 | | 25 | 25 | | 26 | 15 | | 27 | 4 | | 28 | 26 | | 29 | 2 | | 30 | 29 | | 31 | 14 | | 32 | 12 | | 33 | 8 | | 34 | 13 | | 35 | 2 | | 36 | 3 | | 37 | 2 | | 38 | 22 | | 39 | 18 | | 40 | 13 | | 41 | 4 | | 42 | 1 | | 43 | 14 | | 44 | 25 | | 45 | 24 | | 46 | 25 | | 47 | 3 | | 48 | 13 | | 49 | 7 |
| |
| 77.39% | Sentence opener variety | Target: ≥60% unique sentence openers | | consecutiveRepeats | 6 | | diversityRatio | 0.5057471264367817 | | totalSentences | 87 | | uniqueOpeners | 44 | |
| 0.00% | Adverb-first sentence starts | Target: ≥3% sentences starting with an adverb | | adverbCount | 0 | | totalSentences | 74 | | matches | (empty) | | ratio | 0 | |
| 79.46% | Pronoun-first sentence starts | Target: ≤30% sentences starting with a pronoun | | pronounCount | 26 | | totalSentences | 74 | | matches | | 0 | "She'd been watching the Raven's" | | 1 | "Her leather watch pressed cold" | | 2 | "She'd told herself she'd give" | | 3 | "She'd been building a file" | | 4 | "She pushed off the doorframe" | | 5 | "He hadn't looked back yet." | | 6 | "His dark curly hair was" | | 7 | "He looked back." | | 8 | "Their eyes met for half" | | 9 | "She took the alley at" | | 10 | "He vaulted a fallen wheelie" | | 11 | "She reached the top of" | | 12 | "She was off-duty." | | 13 | "She'd called in a favour" | | 14 | "She pressed her hand against" | | 15 | "She'd pulled the board and" | | 16 | "She moved fast down the" | | 17 | "They weren't alarmed." | | 18 | "They were watching, the way" | | 19 | "His left forearm was crossed" |
| | ratio | 0.351 | |
| 54.59% | Subject-first sentence starts | Target: ≤72% sentences starting with a subject | | subjectCount | 60 | | totalSentences | 74 | | matches | | 0 | "The rain came sideways off" | | 1 | "Quinn turned her collar up" | | 2 | "She'd been watching the Raven's" | | 3 | "Her leather watch pressed cold" | | 4 | "She'd told herself she'd give" | | 5 | "She'd been building a file" | | 6 | "Blurry CCTV stills." | | 7 | "A mobile number that went" | | 8 | "Witness statements from people who" | | 9 | "She pushed off the doorframe" | | 10 | "Herrera moved north through Soho," | | 11 | "Quinn kept thirty metres between" | | 12 | "He hadn't looked back yet." | | 13 | "His dark curly hair was" | | 14 | "He looked back." | | 15 | "Their eyes met for half" | | 16 | "Quinn was already moving." | | 17 | "She took the alley at" | | 18 | "Herrera was quick for someone" | | 19 | "He vaulted a fallen wheelie" |
| | ratio | 0.811 | |
| 0.00% | Subordinate conjunction sentence starts | Target: ≥2% sentences starting with a subordinating conjunction | | subConjCount | 0 | | totalSentences | 74 | | matches | (empty) | | ratio | 0 | |
| 94.16% | Technical jargon density | Target: ≤6% sentences with technical-jargon patterns | | analyzedSentences | 44 | | technicalSentenceCount | 3 | | matches | | 0 | "Witness statements from people who described a man matching his description arriving at addresses where, an hour later, someone would be dead or gone or inexpli…" | | 1 | "Herrera moved north through Soho, ducking under scaffolding, cutting through a narrow alley that smelled of bins and old fat from the kitchen extraction fans ov…" | | 2 | "Behind them stood people — and things that were not quite people — selling from arrangements that Quinn's brain kept trying to rationalise and kept failing to." |
| |
| 100.00% | Useless dialogue additions | Target: ≤5% dialogue tags with trailing filler fragments | | totalTags | 3 | | uselessAdditionCount | 0 | | matches | (empty) | |
| 100.00% | Dialogue tag variety (said vs. fancy) | Target: ≤10% fancy dialogue tags | |