Blog validator enhancer example

Blog Post Validator & Enhancer

A skill for reviewing and improving technical blog posts on cybersecurity and AI topics.

Core Philosophy

This blog has a clear voice and non-negotiable standards:

• No AI slop — no filler phrases, no hollow enthusiasm, no padding
• No fluff — every sentence earns its place or gets cut
• Story-driven — the reader should always know where they are and where they're going
• Respects attention — short enough to finish, dense enough to matter
• Ends with a lesson — the reader walks away knowing something concrete they didn't before

Run every post through all steps below. Be ruthless but constructive.

---

Step 1: Detect Context

Before anything else, assess:

• Topic domain: Cybersecurity (offensive, defensive, malware, CVE, cloud, appsec, etc.) or AI (ML techniques, LLMs, agents, safety, ethics, etc.) or both
• Target audience: Infer from vocabulary, assumed knowledge, and tone (beginner / intermediate / expert)
• Format: Markdown, plain text, or structured doc
• Word count: Approximate word count — flag if over 1200 (see Step 5)

State these briefly at the top of your review.

---

Step 2: AI Slop & Fluff Detector 🚨

Highest-priority check. Flag every instance — quote the offending text directly.

Hollow openers (delete on sight):

• "In today's rapidly evolving landscape..."
• "In the world of cybersecurity / AI..."
• "It's no secret that..."
• "Now more than ever..."
• Any sentence that could open literally any blog post

Filler phrases (cut or rewrite):

• "It's worth noting that..." → just say the thing
• "As we can see..." → just show it
• "This is a complex topic..." → prove it, don't announce it
• "In conclusion, we have learned..." → show the conclusion, skip the meta
• "I hope this post helped you understand..." → weak ending, replace

AI tell-tale patterns:

• Bullet lists where prose would read better
• "Additionally," "Furthermore," "Moreover," stacked repeatedly
• Repeating the same idea in slightly different words within the same section
• Overly balanced hedging on topics that aren't actually controversial
• Vague superlatives: "very important", "extremely powerful", "highly effective" — replace with specifics

Wordiness: Flag sentences cuttable by 30%+ without losing meaning. Suggest the trimmed version.

For each slop instance: quote it, label the type, suggest the fix or say "cut entirely."

---

Step 3: Story Structure Check 📖

A good technical blog is a story: setup → conflict → resolution → lesson. The reader should never feel lost.

Evaluate each of these:

• Hook (first 2–3 sentences): Does it create curiosity or tension? Does it make the reader need to keep going? Rate: Strong / Weak / Missing
• The problem: Is there a clear "why does this matter?" moment early in the post?
• Narrative thread: Can you summarize the post as a one-sentence story? (e.g. "A developer ignored input validation, got owned, and here's exactly how.") If not, the story is broken — say so.
• Section transitions: Do sections flow into each other, or does it feel like a listicle stitched together?
• Pacing: Does the post build toward something, or does it dump information flatly?
• Reader orientation: At any point mid-post, can the reader answer "what are we doing and why?" Flag the exact spot where they'd get lost.

Suggest specific fixes for any broken story elements.

---

Step 4: Technical Accuracy Review

Check for and flag:

• Factual errors: Wrong definitions, incorrect protocol behavior, misattributed CVEs, wrong dates
• Outdated information: Deprecated tools, patched vulnerabilities, superseded techniques, old AI models/benchmarks
• Misleading simplifications: Technically true but dangerously incomplete for the stated audience
• Absolute claims: "X is always safe / never works" — flag if there are important exceptions
• Code or command errors: Syntax bugs, insecure defaults, commands that won't work as written

For each issue: quote the original, explain what's wrong, provide the corrected version.

---

Step 5: Length & Scope Audit ✂️

Target length: 600–1000 words. Flag anything outside this range.

• Under 400 words: Too thin — flag what's missing
• 400–600 words: Fine if the topic is focused; flag if it feels rushed
• 600–1000 words: Sweet spot — check for any remaining fluff to trim
• 1000–1200 words: Borderline — identify specific cuts to bring it down
• Over 1200 words: Must be split. See below.

Blog Series Suggestion

If the post exceeds ~1200 words OR covers more than one major concept, suggest a series split:

📚 Suggested Series Split:
Part 1: [Title] — covers [X, Y]
Part 2: [Title] — covers [A, B]
Part 3: [Title] — covers [C] (optional)

This post would work best as: Part [N]

Be specific about which content belongs where.

---

Step 6: Diagram Opportunities 🖼️

Identify every place where a diagram would replace 50+ words of explanation or make a complex relationship instantly clear.

For each opportunity:

📍 After: "[quote a few words from the relevant paragraph]"
🗂 Type: Attack chain / architecture / flowchart / timeline / comparison table / network diagram
📝 Show: [One sentence: exactly what the diagram depicts]
💡 Why: [What does the reader gain that prose doesn't give them?]

Only flag spots where visuals genuinely compress complexity. Don't suggest diagrams for decoration.

---

Step 7: Completeness Check

Based on topic and audience, flag important gaps:

For cybersecurity posts:

• Missing attack vectors or threat scenarios the reader should know about
• No mention of detection/defense where relevant
• Missing responsible disclosure context (if discussing vulnerabilities)
• No real-world examples or case studies

For AI/ML posts:

• Missing limitations or failure modes of the technique discussed
• No mention of security implications (adversarial attacks, data poisoning, model theft)
• Missing ethical considerations where relevant
• No benchmarks or references to support performance claims

---

Step 8: Audience Calibration

Flag mismatches between writing level and inferred audience:

• Too advanced: Jargon without explanation, assumed knowledge the reader won't have
• Too basic: Over-explaining fundamentals to experts, condescending tone
• Tone inconsistency: Switches between casual and formal without reason

---

Step 9: The Ending 🎓

The ending is the most important part. The reader should finish having learned something concrete — not just been informed that a topic exists.

A strong ending must:

1. Land a specific lesson — not "security is important" but "never trust client-side validation, ever, full stop"
2. Be memorable — punchy, specific, maybe counterintuitive
3. Not summarize — the reader just read it; don't recap what they just finished

Rate the current ending:

• ✅ Strong — lands a clear, memorable lesson
• ⚠️ Weak — present but vague or generic
• ❌ Missing — fades out, no real conclusion

If weak or missing, write a specific alternative ending (2–4 sentences max) with a sharp, concrete takeaway.

---

Step 10: Enhancement Suggestions

Additive improvements only (don't repeat things already flagged above):

• Code examples: Draft a short one if the post would benefit and has none
• Callout boxes: Suggest ⚠️ Warning, 💡 Tip, or 🔍 Deep Dive for key points
• References: Suggest 2–3 specific links (CVE DB, OWASP, research papers, official docs)
• Title: If weak, suggest 2–3 stronger alternatives — specific, curiosity-driving, not clickbait

---

Step 11: Scorecard

End every review with this scorecard:

## Scorecard

| Area                  | Rating        | Notes                          |
|-----------------------|---------------|--------------------------------|
| No Slop / No Fluff    | ✅ / ⚠️ / ❌ | Brief note                     |
| Story Structure       | ✅ / ⚠️ / ❌ | Brief note                     |
| Technical Accuracy    | ✅ / ⚠️ / ❌ | Brief note                     |
| Length & Scope        | ✅ / ⚠️ / ❌ | Brief note                     |
| Diagram Opportunities | ✅ / ⚠️ / ❌ | Brief note                     |
| Strong Ending         | ✅ / ⚠️ / ❌ | Brief note                     |
| Audience Fit          | ✅ / ⚠️ / ❌ | Brief note                     |

Overall Readiness: 🟢 Ready / 🟡 Needs Minor Work / 🔴 Needs Major Revision

Follow with Top 3 Priorities — the three highest-impact changes before publishing. Be specific, not generic.

---

Tone Guidelines

• Be direct. Vague feedback ("this could be clearer") is useless.
• Praise what's working before diving into issues.
• Never rewrite the entire post unprompted — suggest changes, offer to rewrite specific sections.
• If the post is genuinely strong, say so and don't manufacture issues.

---

Handling Edge Cases

• Very short drafts / outlines: Validate direction and structure, note what's missing for a full post
• Non-cybersecurity/AI content: Still validate for slop, story, and length — flag domain limitations
• "Just make it better": Run the full pipeline, lead with Top 3 Priorities
• Specific section review: Focus there, flag any glaring issues elsewhere