# Content Quality Audit

> Content Quality Audit

- Skill: `lucadominguez/content-quality-audit` (Agent Skill)
- Install (CLI): `npx skillmds@latest add lucadominguez/content-quality-audit`
- Raw SKILL.md: https://api.skillmd.com/api/skills/lucadominguez/content-quality-audit/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: lucadominguez (https://skillmd.com/u/lucadominguez)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/lucadominguez/content-quality-audit

---

# Content Quality Audit

Score any article against the sirsadalot quality bar. Run before publishing — catches the gaps that separate "good enough" from "someone would pay for this."

## Scoring Rubric (0-100)

### Dimension 1: Citation Density (0-25 points)

| Score | Criteria |
|-------|----------|
| 0-5 | <5 citations, no journal names, no years |
| 6-12 | 5-9 citations, some with years, inconsistent format |
| 13-19 | 10-14 citations, consistent Author (YEAR) Journal format |
| 20-25 | 15+ citations, every factual claim sourced, primary sources, DOI links |

Checklist:
- [ ] Every mechanism claim has a citation
- [ ] Every statistic has a source with year
- [ ] Citations use Author et al. (YEAR), Journal format
- [ ] At least one citation from the last 2 years
- [ ] At least one contrarian/disconfirming paper cited

### Dimension 2: Mechanism Depth (0-25 points)

| Score | Criteria |
|-------|----------|
| 0-5 | Names receptors/pathways without explaining them |
| 6-12 | Explains 2-3 steps in a pathway |
| 13-19 | Walks through full pathway with each molecular step |
| 20-25 | Full pathway + competing hypotheses + "here's what we don't know" + mechanism diagram |

Checklist:
- [ ] Molecular steps are explained in order
- [ ] Each step has a citation
- [ ] Competing mechanism hypotheses mentioned
- [ ] Diagram or visual accompanies the mechanism section
- [ ] Reader can trace the pathway without prior knowledge

### Dimension 3: AI-Tell Score (0-20 points)

Loaded from `humanizer` skill patterns. Score is inverse — higher is better:

| Score | Meaning |
|-------|---------|
| 0-5 | Heavy AI patterns: "crucial role," "rapidly evolving," uniform sentence length, no opinions |
| 6-12 | Some AI patterns, mostly clean but lacks texture |
| 13-19 | Mostly clean, 1-2 tells remain, decent burstiness |
| 20 | No detectable AI patterns, strong voice, varied rhythm |

Quick scan checklist:
- [ ] No "in today's rapidly evolving landscape"
- [ ] No "plays a crucial role in"
- [ ] No "further research is needed"
- [ ] No "-ing phrase, main clause" tack-on pattern
- [ ] Sentence length varies (2-word to 35-word range)
- [ ] At least one opinionated statement
- [ ] At least one "I don't know" or uncertainty admission
- [ ] No bolded key terms (if used, flagged)

### Dimension 4: Information Density (0-15 points)

Per-paragraph audit. Each paragraph must contain at least one of:

- A specific number with units
- A citation with journal and year
- A mechanism name (receptor, kinase, pathway)
- A comparative statistic
- A dollar amount or cost figure
- A named risk or diagnostic criterion

| Score | Criteria |
|-------|----------|
| 0-5 | >40% of paragraphs lack a concrete element |
| 6-10 | 20-40% lack concrete elements |
| 11-15 | <20% lack concrete elements |

### Dimension 5: Visual Completeness (0-10 points)

| Visuals present | Score |
|----------------|-------|
| 0 visuals | 0 |
| 1 visual (any type) | 3 |
| 2 different visual types | 6 |
| 3+ different visual types | 10 |

Visual types: comparison chart, mechanism diagram, protocol table, dose-response curve, forest plot, infographic, concept diagram, excalidraw map.

### Dimension 6: Structural Completeness (0-5 points)

sirsadalot structure check:

- [ ] Bottom Line box or TLDR at the top (1 pt)
- [ ] Context/why-this-matters section (1 pt)
- [ ] Mechanism deep-dive section (1 pt)
- [ ] Practical protocols/dosing section (1 pt)
- [ ] Further Reading with annotated papers (1 pt)

---

## How to Run the Audit

```bash
# Point it at your article
python3 << 'PYEOF'
article_path = '/home/lenovo/writing/ketamine-mechanisms-2026/final.md'
with open(article_path) as f:
    text = f.read()

# Count citations
import re
citations = re.findall(r'\((\d{4})\)', text)  # years in parens
journals = re.findall(r'\*(Biological|Nature|Science|Neuron|Journal|Archives|American)[^*]*\*', text)
print(f"Unique citation years: {len(set(citations))}")
print(f"Journal mentions: {len(journals)}")

# Count paragraphs with concrete elements
paragraphs = [p for p in text.split('\n\n') if len(p) > 100]
has_number = sum(1 for p in paragraphs if re.search(r'\d+%|\d+ mg|\d+\.\d+', p))
has_citation = sum(1 for p in paragraphs if re.search(r'\(\d{4}\)', p))
print(f"Paragraphs: {len(paragraphs)}")
print(f"With numbers: {has_number} ({100*has_number//max(len(paragraphs),1)}%)")
print(f"With citations: {has_citation} ({100*has_citation//max(len(paragraphs),1)}%)")

# Count visuals referenced
visuals = re.findall(r'!\[.*?\]\(.*?\)|chart|diagram|figure|infographic|table', text, re.IGNORECASE)
print(f"Visual references: {len(visuals)}")

# AI tell scan
ai_tells = ['crucial role', 'rapidly evolving', 'in today's', 'further research is needed',
            'plays a key role', 'it is important to note', 'has been shown to']
for tell in ai_tells:
    if tell.lower() in text.lower():
        print(f"AI TELL FOUND: '{tell}'")

# Structure check
structure_checks = {
    'Bottom Line / TLDR': any(x in text[:500].lower() for x in ['bottom line', 'tldr', 'verdict', 'takeaway']),
    'Context section': 'context' in text.lower() or 'why this matters' in text.lower(),
    'Mechanism deep-dive': 'mechanism' in text.lower() and 'pathway' in text.lower(),
    'Protocol section': 'protocol' in text.lower() or 'dose' in text.lower() or 'dosing' in text.lower(),
    'Further Reading': 'further reading' in text.lower() or 'references' in text.lower(),
}
for check, passed in structure_checks.items():
    print(f"{'[✓]' if passed else '[ ]'} {check}")
PYEOF
```

## Scorecard Output Template

```
╔══════════════════════════════════════╗
║     CONTENT QUALITY AUDIT           ║
║     Article: [title]                ║
╠══════════════════════════════════════╣
║                                      ║
║  Citation Density:     XX/25  [bar]  ║
║  Mechanism Depth:      XX/25  [bar]  ║
║  AI-Tell Score:        XX/20  [bar]  ║
║  Information Density:  XX/15  [bar]  ║
║  Visual Completeness:  XX/10  [bar]  ║
║  Structure:            XX/5   [bar]  ║
║                            ─────     ║
║  TOTAL:                XX/100        ║
║                                      ║
║  sirsadalot threshold: 85+           ║
║  Publishable:          70+           ║
║  Needs work:           <70           ║
╚══════════════════════════════════════╝

FIXES REQUIRED:
1. [Most impactful gap]
2. [Second]
3. [Third]
```

## PITFALLS

1. **Don't score before the humanization pass.** Running the audit on a first draft will give artificially low AI-tell scores. Audit the final version.
2. **Citation counting is mechanical but meaningful.** Regex can't tell if a citation is real or hallucinated. Spot-check 3 citations manually.
3. **Information density varies by section.** A "Further Reading" section will naturally have lower density than the mechanism section. Weight the mechanism and evidence sections more heavily.