| 100.00% | Adverbs in dialogue tags | Target: ≤10% dialogue tags with adverbs | | totalTags | 10 | | adverbTagCount | 2 | | adverbTags | | 0 | "The words came automatically [automatically]" | | 1 | "Quinn said slowly [slowly]" |
| | dialogueSentences | 51 | | tagDensity | 0.196 | | leniency | 0.392 | | rawRatio | 0.2 | | effectiveRatio | 0.078 | |
| 95.81% | AI-ism adverb frequency | Target: <2% AI-ism adverbs (58 tracked) | | wordCount | 1193 | | totalAiIsmAdverbs | 1 | | found | | | highlights | | |
| 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) | |
| 49.71% | AI-ism word frequency | Target: <2% AI-ism words (290 tracked) | | wordCount | 1193 | | totalAiIsms | 12 | | found | | | highlights | | 0 | "familiar" | | 1 | "weight" | | 2 | "methodical" | | 3 | "stomach" | | 4 | "quivered" | | 5 | "standard" | | 6 | "database" | | 7 | "firmly" | | 8 | "furrowed" | | 9 | "tracing" | | 10 | "etched" |
| |
| 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 | 78 | | matches | (empty) | |
| 100.00% | Filter word density | Target: ≤3% sentences with filter/hedge words | | filterCount | 1 | | hedgeCount | 1 | | narrationSentences | 78 | | filterMatches | | | hedgeMatches | | |
| 100.00% | Gibberish response detection | Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words) | | analyzedSentences | 118 | | gibberishSentences | 0 | | adjustedGibberishSentences | 0 | | longSentenceCount | 0 | | runOnParagraphCount | 0 | | giantParagraphCount | 0 | | wordSaladCount | 0 | | repetitionLoopCount | 0 | | controlTokenCount | 0 | | maxSentenceWordsSeen | 33 | | ratio | 0 | | matches | (empty) | |
| 100.00% | Markdown formatting overuse | Target: ≤5% words in markdown formatting | | markdownSpans | 1 | | markdownWords | 13 | | totalWords | 1189 | | ratio | 0.011 | | matches | | 0 | "She found the Market. Ask yourself who didn't want her to come back." |
| |
| 100.00% | Missing dialogue indicators (quotation marks) | Target: ≤10% speech attributions without quotation marks | | totalAttributions | 6 | | unquotedAttributions | 0 | | matches | (empty) | |
| 67.26% | Name drop frequency | Target: ≤1.0 per-name mentions per 100 words | | totalMentions | 34 | | wordCount | 846 | | uniqueNames | 14 | | maxNameDensity | 1.65 | | worstName | "Quinn" | | maxWindowNameDensity | 2.5 | | worstWindowName | "Quinn" | | discoveredNames | | Platform | 2 | | Harlow | 1 | | Quinn | 14 | | Davies | 6 | | Morris | 2 | | Town | 1 | | Tuesday | 1 | | Seven | 1 | | Deptford | 1 | | London | 1 | | Underground | 1 | | Eva | 1 | | Kowalski | 1 | | Market | 1 |
| | persons | | 0 | "Harlow" | | 1 | "Quinn" | | 2 | "Davies" | | 3 | "Morris" | | 4 | "Seven" | | 5 | "Underground" | | 6 | "Eva" | | 7 | "Kowalski" |
| | places | | 0 | "Platform" | | 1 | "Town" | | 2 | "Deptford" | | 3 | "London" |
| | globalScore | 0.673 | | windowScore | 0.833 | |
| 18.42% | Narrator intent-glossing | Target: ≤2% narration sentences with intent-glossing patterns | | analyzedSentences | 57 | | glossingSentenceCount | 3 | | matches | | 0 | "as if reaching for a train that would never come" | | 1 | "orpse that had apparently been lying undiscov" | | 2 | "felt like they were waiting for somethi" |
| |
| 100.00% | "Not X but Y" pattern overuse | Target: ≤1 "not X but Y" per 1000 words | | totalMatches | 0 | | per1kWords | 0 | | wordCount | 1189 | | matches | (empty) | |
| 100.00% | Overuse of "that" (subordinate clause padding) | Target: ≤2% sentences with "that" clauses | | thatCount | 0 | | totalSentences | 118 | | matches | (empty) | |
| 100.00% | Paragraph length variance | Target: CV ≥0.5 for paragraph word counts | | totalParagraphs | 63 | | mean | 18.87 | | std | 17.42 | | cv | 0.923 | | sampleLengths | | 0 | 22 | | 1 | 59 | | 2 | 20 | | 3 | 1 | | 4 | 16 | | 5 | 46 | | 6 | 1 | | 7 | 9 | | 8 | 11 | | 9 | 3 | | 10 | 2 | | 11 | 10 | | 12 | 67 | | 13 | 3 | | 14 | 28 | | 15 | 42 | | 16 | 4 | | 17 | 3 | | 18 | 2 | | 19 | 12 | | 20 | 36 | | 21 | 28 | | 22 | 12 | | 23 | 2 | | 24 | 7 | | 25 | 42 | | 26 | 12 | | 27 | 9 | | 28 | 1 | | 29 | 12 | | 30 | 34 | | 31 | 51 | | 32 | 12 | | 33 | 9 | | 34 | 9 | | 35 | 2 | | 36 | 9 | | 37 | 36 | | 38 | 1 | | 39 | 1 | | 40 | 33 | | 41 | 6 | | 42 | 33 | | 43 | 9 | | 44 | 57 | | 45 | 18 | | 46 | 27 | | 47 | 3 | | 48 | 11 | | 49 | 38 |
| |
| 100.00% | Passive voice overuse | Target: ≤2% passive sentences | | passiveCount | 1 | | totalSentences | 78 | | matches | | |
| 48.48% | Past progressive (was/were + -ing) overuse | Target: ≤2% past progressive verbs | | pastProgressiveCount | 3 | | totalVerbs | 132 | | matches | | 0 | "was examining" | | 1 | "were waiting" | | 2 | "was giving" |
| |
| 100.00% | Em-dash & semicolon overuse | Target: ≤2% sentences with em-dashes/semicolons | | emDashCount | 0 | | semicolonCount | 0 | | flaggedSentences | 0 | | totalSentences | 118 | | ratio | 0 | | matches | (empty) | |
| 94.00% | Purple prose (modifier overload) | Target: <4% adverbs, <2% -ly adverbs, no adj stacking | | wordCount | 848 | | adjectiveStacks | 1 | | stackExamples | | | adverbCount | 27 | | adverbRatio | 0.031839622641509434 | | lyAdverbCount | 11 | | lyAdverbRatio | 0.012971698113207548 | |
| 100.00% | Repeated phrase echo | Target: ≤20% sentences with echoes (window: 2) | | totalSentences | 118 | | echoCount | 0 | | echoWords | (empty) | |
| 100.00% | Sentence length variance | Target: CV ≥0.4 for sentence word counts | | totalSentences | 118 | | mean | 10.08 | | std | 7.32 | | cv | 0.726 | | sampleLengths | | 0 | 22 | | 1 | 18 | | 2 | 15 | | 3 | 8 | | 4 | 8 | | 5 | 10 | | 6 | 12 | | 7 | 8 | | 8 | 1 | | 9 | 16 | | 10 | 19 | | 11 | 12 | | 12 | 2 | | 13 | 13 | | 14 | 1 | | 15 | 9 | | 16 | 8 | | 17 | 3 | | 18 | 3 | | 19 | 2 | | 20 | 5 | | 21 | 2 | | 22 | 3 | | 23 | 4 | | 24 | 16 | | 25 | 31 | | 26 | 16 | | 27 | 3 | | 28 | 17 | | 29 | 11 | | 30 | 7 | | 31 | 17 | | 32 | 18 | | 33 | 4 | | 34 | 3 | | 35 | 2 | | 36 | 6 | | 37 | 6 | | 38 | 11 | | 39 | 15 | | 40 | 10 | | 41 | 7 | | 42 | 15 | | 43 | 6 | | 44 | 12 | | 45 | 2 | | 46 | 7 | | 47 | 9 | | 48 | 4 | | 49 | 10 |
| |
| 89.83% | Sentence opener variety | Target: ≥60% unique sentence openers | | consecutiveRepeats | 4 | | diversityRatio | 0.559322033898305 | | totalSentences | 118 | | uniqueOpeners | 66 | |
| 91.32% | Adverb-first sentence starts | Target: ≥3% sentences starting with an adverb | | adverbCount | 2 | | totalSentences | 73 | | matches | | 0 | "Somewhere that had cost Morris" | | 1 | "Too calm for a man" |
| | ratio | 0.027 | |
| 100.00% | Pronoun-first sentence starts | Target: ≤30% sentences starting with a pronoun | | pronounCount | 15 | | totalSentences | 73 | | matches | | 0 | "She didn't believe in luck." | | 1 | "She surveyed the platform with" | | 2 | "She read them off, her" | | 3 | "She stood too fast, blood" | | 4 | "She photographed the scene with" | | 5 | "He held up a small" | | 6 | "It fit in his palm" | | 7 | "It was warm, which was" | | 8 | "She'd held a thousand pieces" | | 9 | "she said, her voice steady" | | 10 | "She met his eyes, and" | | 11 | "She stripped off her gloves," | | 12 | "Her phone buzzed." | | 13 | "*She found the Market." | | 14 | "His hands were steady, his" |
| | ratio | 0.205 | |
| 90.14% | Subject-first sentence starts | Target: ≤72% sentences starting with a subject | | subjectCount | 54 | | totalSentences | 73 | | matches | | 0 | "The body lay sprawled across" | | 1 | "Detective Harlow Quinn ducked under" | | 2 | "This one couldn't have been" | | 3 | "PC Davies nodded from his" | | 4 | "Quinn crouched beside the body," | | 5 | "The victim's satchel lay open" | | 6 | "The kind with cracked spines" | | 7 | "Quinn's fingers paused over the" | | 8 | "She didn't believe in luck." | | 9 | "The platform was wrong." | | 10 | "That was the first thing" | | 11 | "Camden Town station saw thousands" | | 12 | "The CCTV footage would tell" | | 13 | "Davies consulted his notebook" | | 14 | "Quinn stood, her knees protesting" | | 15 | "She surveyed the platform with" | | 16 | "Davies squinted at the scattered" | | 17 | "Quinn pointed to the pattern" | | 18 | "She read them off, her" | | 19 | "Davies had the grace to" |
| | ratio | 0.74 | |
| 0.00% | Subordinate conjunction sentence starts | Target: ≥2% sentences starting with a subordinating conjunction | | subConjCount | 0 | | totalSentences | 73 | | matches | (empty) | | ratio | 0 | |
| 3.97% | Technical jargon density | Target: ≤6% sentences with technical-jargon patterns | | analyzedSentences | 36 | | technicalSentenceCount | 7 | | matches | | 0 | "Camden Town station saw thousands of commuters daily, yet here they were at half past two on a Tuesday afternoon with a corpse that had apparently been lying un…" | | 1 | "She surveyed the platform with the methodical precision that had earned her three commendations and one disciplinary hearing." | | 2 | "The compass was old, its casing wearing a patina of verdigris that spoke of decades, maybe centuries." | | 3 | "Quinn turned back to the body, to Eva Kowalski with her red curls and her ancient books and the cuts on her palm that matched the ones she still saw in her nigh…" | | 4 | "The compass, which had stopped spinning and now pointed firmly at a spot on the wall where the mortar between two bricks looked fractionally newer than its neig…" | | 5 | "Unknown number, which meant it had been routed through enough proxies to make tracing impossible." | | 6 | "Across the platform, the maintenance worker who'd discovered the body was giving his statement to a uniformed officer." |
| |
| 25.00% | Useless dialogue additions | Target: ≤5% dialogue tags with trailing filler fragments | | totalTags | 10 | | uselessAdditionCount | 2 | | matches | | 0 | "she said, her voice steady despite the crawling sensation up her spine" | | 1 | "She met, and whatever he saw in hers made him straighten" |
| |
| 100.00% | Dialogue tag variety (said vs. fancy) | Target: ≤10% fancy dialogue tags | | totalTags | 3 | | fancyCount | 0 | | fancyTags | (empty) | | dialogueSentences | 51 | | tagDensity | 0.059 | | leniency | 0.118 | | rawRatio | 0 | | effectiveRatio | 0 | |