| 100.00% | Adverbs in dialogue tags | Target: ≤10% dialogue tags with adverbs | | totalTags | 9 | | adverbTagCount | 0 | | adverbTags | (empty) | | dialogueSentences | 16 | | tagDensity | 0.563 | | leniency | 1 | | rawRatio | 0 | | effectiveRatio | 0 | |
| 82.48% | AI-ism adverb frequency | Target: <2% AI-ism adverbs (58 tracked) | | wordCount | 856 | | totalAiIsmAdverbs | 3 | | found | | 0 | | adverb | "deliberately" | | count | 1 |
| | 1 | |
| | highlights | | 0 | "deliberately" | | 1 | "suddenly" |
| |
| 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) | |
| 47.43% | AI-ism word frequency | Target: <2% AI-ism words (290 tracked) | | wordCount | 856 | | totalAiIsms | 9 | | found | | | highlights | | 0 | "loomed" | | 1 | "stomach" | | 2 | "maw" | | 3 | "facade" | | 4 | "dancing" | | 5 | "gloom" | | 6 | "scanning" | | 7 | "raced" |
| |
| 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 | 65 | | matches | (empty) | |
| 98.90% | Filter word density | Target: ≤3% sentences with filter/hedge words | | filterCount | 1 | | hedgeCount | 1 | | narrationSentences | 65 | | filterMatches | | | hedgeMatches | | |
| 100.00% | Gibberish response detection | Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words) | | analyzedSentences | 72 | | gibberishSentences | 0 | | adjustedGibberishSentences | 0 | | longSentenceCount | 0 | | runOnParagraphCount | 0 | | giantParagraphCount | 0 | | wordSaladCount | 0 | | repetitionLoopCount | 0 | | controlTokenCount | 0 | | maxSentenceWordsSeen | 30 | | ratio | 0 | | matches | (empty) | |
| 100.00% | Markdown formatting overuse | Target: ≤5% words in markdown formatting | | markdownSpans | 2 | | markdownWords | 10 | | totalWords | 848 | | ratio | 0.012 | | matches | | 0 | "Evidence doesn’t add up. Meet me at the Veil." | | 1 | "down" |
| |
| 100.00% | Missing dialogue indicators (quotation marks) | Target: ≤10% speech attributions without quotation marks | | totalAttributions | 8 | | unquotedAttributions | 0 | | matches | (empty) | |
| 66.67% | Name drop frequency | Target: ≤1.0 per-name mentions per 100 words | | totalMentions | 47 | | wordCount | 751 | | uniqueNames | 16 | | maxNameDensity | 1.46 | | worstName | "Quinn" | | maxWindowNameDensity | 3 | | worstWindowName | "Quinn" | | discoveredNames | | HarlowQuinn | 1 | | Camden | 2 | | High | 1 | | Street | 1 | | Tube | 3 | | Metropolitan | 1 | | Police | 1 | | Veil | 7 | | Market | 4 | | Quinn | 11 | | Kowalski | 1 | | Greek | 1 | | Eva | 10 | | Ouroboros | 1 | | Compass | 1 | | Morris | 1 |
| | persons | | 0 | "HarlowQuinn" | | 1 | "Police" | | 2 | "Quinn" | | 3 | "Kowalski" | | 4 | "Eva" | | 5 | "Compass" | | 6 | "Morris" |
| | places | | 0 | "Camden" | | 1 | "High" | | 2 | "Street" | | 3 | "Veil" |
| | globalScore | 0.768 | | windowScore | 0.667 | |
| 0.00% | Narrator intent-glossing | Target: ≤2% narration sentences with intent-glossing patterns | | analyzedSentences | 46 | | glossingSentenceCount | 4 | | matches | | 0 | "shadows that seemed to writhe across the peeling posters advertising long-dead services" | | 1 | "looked like ancient Greek script" | | 2 | "looked like it had been pulled from a cha" | | 3 | "looked like a crude symbol, half-erased b" |
| |
| 100.00% | "Not X but Y" pattern overuse | Target: ≤1 "not X but Y" per 1000 words | | totalMatches | 0 | | per1kWords | 0 | | wordCount | 848 | | matches | (empty) | |
| 100.00% | Overuse of "that" (subordinate clause padding) | Target: ≤2% sentences with "that" clauses | | thatCount | 0 | | totalSentences | 72 | | matches | (empty) | |
| 100.00% | Paragraph length variance | Target: CV ≥0.5 for paragraph word counts | | totalParagraphs | 14 | | mean | 60.57 | | std | 33.1 | | cv | 0.547 | | sampleLengths | | 0 | 123 | | 1 | 96 | | 2 | 84 | | 3 | 17 | | 4 | 98 | | 5 | 54 | | 6 | 46 | | 7 | 43 | | 8 | 15 | | 9 | 59 | | 10 | 80 | | 11 | 10 | | 12 | 42 | | 13 | 81 |
| |
| 83.67% | Passive voice overuse | Target: ≤2% passive sentences | | passiveCount | 4 | | totalSentences | 65 | | matches | | 0 | "been summoned" | | 1 | "been pulled" | | 2 | "was positioned" | | 3 | "was known" |
| |
| 0.00% | Past progressive (was/were + -ing) overuse | Target: ≤2% past progressive verbs | | pastProgressiveCount | 4 | | totalVerbs | 101 | | matches | | 0 | "were looking" | | 1 | "wasn’t pointing" | | 2 | "was pointing" | | 3 | "was pointing" |
| |
| 0.00% | Em-dash & semicolon overuse | Target: ≤2% sentences with em-dashes/semicolons | | emDashCount | 4 | | semicolonCount | 2 | | flaggedSentences | 5 | | totalSentences | 72 | | ratio | 0.069 | | matches | | 0 | "This wasn’t just a crime scene; it was a trap." | | 1 | "The air hung thick with the scent of mildew and something else—something ancient and faintly sweet, like decaying flowers." | | 2 | "The cause of death was obvious—a single, precise stab wound to the heart." | | 3 | "And this location—an abandoned Tube station beneath Camden—was a known entry point." | | 4 | "The Veil Market wasn’t just a market; it was a gateway." |
| |
| 88.00% | Purple prose (modifier overload) | Target: <4% adverbs, <2% -ly adverbs, no adj stacking | | wordCount | 759 | | adjectiveStacks | 2 | | stackExamples | | 0 | "lay open beside her," | | 1 | "cheap, ill-fitting coat" |
| | adverbCount | 25 | | adverbRatio | 0.03293807641633729 | | lyAdverbCount | 10 | | lyAdverbRatio | 0.013175230566534914 | |
| 100.00% | Repeated phrase echo | Target: ≤20% sentences with echoes (window: 2) | | totalSentences | 72 | | echoCount | 0 | | echoWords | (empty) | |
| 100.00% | Sentence length variance | Target: CV ≥0.4 for sentence word counts | | totalSentences | 72 | | mean | 11.78 | | std | 7.75 | | cv | 0.658 | | sampleLengths | | 0 | 28 | | 1 | 24 | | 2 | 12 | | 3 | 27 | | 4 | 22 | | 5 | 10 | | 6 | 24 | | 7 | 20 | | 8 | 13 | | 9 | 20 | | 10 | 19 | | 11 | 19 | | 12 | 24 | | 13 | 26 | | 14 | 15 | | 15 | 7 | | 16 | 10 | | 17 | 6 | | 18 | 22 | | 19 | 9 | | 20 | 13 | | 21 | 5 | | 22 | 15 | | 23 | 7 | | 24 | 2 | | 25 | 3 | | 26 | 16 | | 27 | 10 | | 28 | 7 | | 29 | 2 | | 30 | 5 | | 31 | 3 | | 32 | 3 | | 33 | 12 | | 34 | 12 | | 35 | 8 | | 36 | 18 | | 37 | 11 | | 38 | 9 | | 39 | 10 | | 40 | 24 | | 41 | 6 | | 42 | 3 | | 43 | 9 | | 44 | 6 | | 45 | 9 | | 46 | 19 | | 47 | 23 | | 48 | 8 | | 49 | 6 |
| |
| 50.46% | Sentence opener variety | Target: ≥60% unique sentence openers | | consecutiveRepeats | 7 | | diversityRatio | 0.3611111111111111 | | totalSentences | 72 | | uniqueOpeners | 26 | |
| 54.64% | Adverb-first sentence starts | Target: ≥3% sentences starting with an adverb | | adverbCount | 1 | | totalSentences | 61 | | matches | | 0 | "Just the woman, dead, and" |
| | ratio | 0.016 | |
| 100.00% | Pronoun-first sentence starts | Target: ≤30% sentences starting with a pronoun | | pronounCount | 14 | | totalSentences | 61 | | matches | | 0 | "She’d been summoned here by" | | 1 | "She’d never been here before," | | 2 | "Her eyes were wide, frozen" | | 3 | "She pointed to a faint," | | 4 | "It looked like a crude" | | 5 | "It was a simple circle" | | 6 | "She gestured towards a patch" | | 7 | "She’d never used it, never" | | 8 | "It wasn’t pointing north." | | 9 | "It was pointing *down*." | | 10 | "Her partner, DS Morris, had" | | 11 | "She’d never understood it, never" | | 12 | "It was a message." | | 13 | "She had to follow it." |
| | ratio | 0.23 | |
| 74.75% | Subject-first sentence starts | Target: ≤72% sentences starting with a subject | | subjectCount | 47 | | totalSentences | 61 | | matches | | 0 | "Detective HarlowQuinn’s boots scraped gravel" | | 1 | "The abandoned Tube station loomed" | | 2 | "She’d been summoned here by" | | 3 | "She’d never been here before," | | 4 | "This wasn’t just a crime" | | 5 | "The station’s entrance was a" | | 6 | "Quinn pushed through the rusted" | | 7 | "A single, flickering fluorescent tube" | | 8 | "The air hung thick with" | | 9 | "Detective Quinn’s flashlight beam cut" | | 10 | "Eva Kowalski, her red hair" | | 11 | "Eva’s round glasses glinted in" | | 12 | "Eva jumped as the light" | | 13 | "Eva said, her voice tight" | | 14 | "Quinn’s flashlight swept over the" | | 15 | "A young woman, maybe twenty-five," | | 16 | "Her eyes were wide, frozen" | | 17 | "The cause of death was" | | 18 | "The body was positioned oddly," | | 19 | "Quinn asked, her voice low," |
| | ratio | 0.77 | |
| 0.00% | Subordinate conjunction sentence starts | Target: ≥2% sentences starting with a subordinating conjunction | | subConjCount | 0 | | totalSentences | 61 | | matches | (empty) | | ratio | 0 | |
| 71.43% | Technical jargon density | Target: ≤6% sentences with technical-jargon patterns | | analyzedSentences | 30 | | technicalSentenceCount | 3 | | matches | | 0 | "Meet me at the Veil.* The bone token in her pocket felt cold and heavy, a key to this underground labyrinth that moved locations every full moon." | | 1 | "A single, flickering fluorescent tube cast long, dancing shadows that seemed to writhe across the peeling posters advertising long-dead services." | | 2 | "A young woman, maybe twenty-five, dressed in a cheap, ill-fitting coat that looked like it had been pulled from a charity bin." |
| |
| 0.00% | Useless dialogue additions | Target: ≤5% dialogue tags with trailing filler fragments | | totalTags | 9 | | uselessAdditionCount | 5 | | matches | | 0 | "Eva said, her voice tight" | | 1 | "Quinn asked, her voice low, scanning the shadows" | | 2 | "Quinn said, her gaze sweeping the platform again" | | 3 | "Eva said, her voice dropping" | | 4 | "Quinn breathed, her voice barely audible" |
| |
| 25.00% | Dialogue tag variety (said vs. fancy) | Target: ≤10% fancy dialogue tags | | totalTags | 6 | | fancyCount | 2 | | fancyTags | | 0 | "Eva murmured (murmur)" | | 1 | "Quinn breathed (breathe)" |
| | dialogueSentences | 16 | | tagDensity | 0.375 | | leniency | 0.75 | | rawRatio | 0.333 | | effectiveRatio | 0.25 | |