| name | evaluate-content |
| description | Use when judging content quality OR editing/improving existing copy: shareability, readability, voice, cuttability, angle, copy sweeps. |
Evaluate Content
Honest, brutal content evaluation. Run this on anything before it goes live, articles, threads, tweets, headlines, outlines.
When to Use Which Framework
| Task | Framework |
|---|
| Evaluating content quality | The Six Questions |
| Editing/improving existing copy | The Seven Sweeps |
| Quick polish pass | Quick-Pass Editing Checks |
| Monthly content review | Pillar-Level Evaluation |
| Video hooks | Short-Form Video Hook Checklist |
The Six Questions
Every piece of content gets judged on these six questions. Score each 1-5 and explain why.
1. "Would someone share this in a group chat?"
What you're really asking: Is this interesting enough that someone would text it to a friend unprompted?
Score 5: "I'm sending this to three people right now." It has a surprising fact, a spicy take, or it articulates something people feel but haven't been able to say.
Score 3: "Hm, interesting." They'd read it but not forward it. It's informative but not remarkable.
Score 1: "Why would I share this?" Generic, nothing new, reads like a summary of things people already know.
Tests:
- The Signulll test (John Rush): Would someone share this to virtue signal, say something they can't say themselves, or show off their taste?
- The screenshot test: Would someone screenshot a specific line and post it?
- The "actually" test: Does it contain something that makes people say "actually, did you know..."?
Common failures:
- Too balanced, no point of view worth sharing
- Too obvious, says what everyone already thinks
- Too niche, interesting but the reader can't explain it to someone else in one sentence
2. "Is this informative or fun to read?"
What you're really asking: Does the reader get something from this, either knowledge they didn't have, or entertainment they enjoyed?
Score 5: Reader learned something specific they can use, OR they genuinely enjoyed reading it. Ideally both.
Score 3: Some useful info but presented in a way that's easy to skim-and-forget. Or mildly entertaining but no substance.
Score 1: Neither. It's a slog to get through and you don't learn anything new.
Tests:
- The "so what?" test: After every section, ask "so what?" If you can't answer in one sentence, the section failed.
- The retention test: Close the article. What do you remember? If the answer is "not much", it's not informative enough.
- The re-read test: Would you read any part of this twice? If yes, which part? That's your best content.
Common failures:
- All setup, no payoff, spends 500 words explaining context before saying anything new
- Data without insight, shows numbers but doesn't say what they mean
- Informative but dry, reads like a textbook, technically correct but boring
3. "Does it write like someone is talking?"
What you're really asking: Does this sound like a human wrote it, or does it sound like AI / a corporate blog / a college essay?
Score 5: You can hear the author's voice. Irregular sentences. Personal asides. Opinions stated directly. You'd recognize this writer's style blind.
Score 3: Mostly natural but has some stiff patches. Occasional "it's important to note" or "in this article we'll explore" slipping through.
Score 1: Full robot. Uniform sentence length. No personality. Could've been written by any AI or any corporate comms team.
Tests:
- Read it aloud. Where do you stumble? Those sentences are broken.
- The "bar test": Would you say this to a friend at a bar? If it sounds weird spoken, rewrite it.
- Ctrl+F the Kill Phrases list in the
writer skill's references/WRITING-STYLE.md. Any hits = automatic deduction. This is the ground truth for voice, not generic "sounds human" criteria.
- Check for first-person. Zero "I" in an opinion piece = probably slop.
- Check sentence variety. Same-length sentences in a row = robot cadence.
- Cross-check against the Voice Fingerprint section of the
writer skill's references/WRITING-STYLE.md: does it lead with a concrete moment (not a thesis)? Does it use specific numbers? Does it show the work rather than claim it? If none of these are present, it's not Eric's voice.
Common failures:
- Every paragraph starts with "This" or "The"
- No personality, could be attributed to anyone
- Uses filler transitions: "Furthermore," "Moreover," "Additionally,"
- Three adjectives in a row: "powerful, innovative, and transformative"
4. "What could I cut out?"
What you're really asking: Where does the reader's attention drift? What's there for the writer, not the reader?
Score 5 (tight): Every sentence earns its place. You couldn't cut anything without losing meaning.
Score 3 (some fat): 10-20% could go. Some repeated points, some throat-clearing, some sections that don't advance the argument.
Score 1 (bloated): 30%+ is filler. Repeats the same point in different words. Long preambles. Obvious padding.
How to evaluate:
- Write with your eraser (word-level pass). Before marking sentences, go word by word and try to delete each one. Every word the sentence survives without makes what's left stronger, so it goes. This is a distinct pass from cutting whole sentences below: a sentence can be ESSENTIAL and still carry three deletable words inside it. Hunt qualifiers, intensifiers, hedges, and any word doing no work.
- Mark every sentence as ESSENTIAL, NICE-TO-HAVE, or CUT
- Look for repeated points stated differently (the most common padding)
- Check opening paragraphs of each section, are the first 1-2 sentences throat-clearing?
- Count qualifiers: "very," "really," "quite," "somewhat," "in order to", each one is a candidate for deletion
- If a section could be summarized in one sentence and nothing would be lost, cut it to one sentence
Give a specific cut list:
CUT: [paragraph/sentence] / reason
CUT: [paragraph/sentence] / reason
TRIM: [paragraph/sentence] -> [shorter version]
If there's genuinely nothing to cut, write "CUT: nothing" and move on. Don't invent a
marginal cut to fill the section. A piece where every sentence earns its place is what a
5 looks like, so an empty cut list is the correct output, not a missing one.
5. "What's the unique perspective or driving emotion?"
What you're really asking: Why does THIS person need to write THIS article? What's the angle only they can bring?
Score 5: Crystal clear point of view. You know exactly what the author believes and why. There's a driving emotion, frustration with the status quo, excitement about a discovery, pride in something they built.
Score 3: Has an angle but doesn't commit to it fully. Hedges too much. Or the emotion is there but buried under too much analysis.
Score 1: No angle. This could be a Wikipedia summary. No emotion. No stakes. No reason this person specifically needed to write this.
Identify:
- The thesis in one sentence. If you can't state it in one sentence, the article doesn't have one.
- The driving emotion: frustration, relief, pride, curiosity, anger, surprise? Name it.
- The "only I" factor: What does this author know/have experienced that makes them uniquely qualified? If anyone could've written this, it's not differentiated.
Common failures:
- "Balanced" pieces with no thesis, presenting information without a point of view
- Borrowed opinions, restating what everyone else says without adding anything
- The emotion exists but is buried in paragraph 7 instead of leading
6. "What kind of person would be looking for this article?"
What you're really asking: Who is this for? Can you picture them? What did they Google or scroll past that led them here?
Score 5: You can describe the reader in one sentence. Their problem is clear. The article directly addresses their situation.
Score 3: General audience. "People interested in investing." Not wrong, but not specific enough to drive strong resonance.
Score 1: Nobody in particular. Or the article serves two different audiences and does neither well.
Identify:
- The reader in one sentence: "A 28-year-old PM who just got assigned to an AI product and has no idea what evals are."
- Their trigger: What happened that made them need this? "Their AI shipped a wrong answer and the CEO noticed."
- What they'd Google: The search query that would lead to this. "how to test AI product quality"
- Their awareness level (from hooks skill):
- Unaware: doesn't know the problem exists
- Problem-aware: knows the problem, doesn't know the solution
- Solution-aware: knows solutions exist, comparing options
- Product-aware: knows your specific product
Company Thesis / Category Essay Reviews
When reviewing company blog posts, founder essays, market theses, or category-design pieces, do not stop at prose quality. Judge whether the piece makes the company feel strategically inevitable.
Use references/company-thesis-review.md when the user asks whether an article "has everything," shares multiple related company posts, or wants strategic feedback on a market thesis. The key checks are: thesis completeness, first product wedge, trust machinery, buyer/allocator POV, workflow concreteness, category boundaries, proof quality, and prediction hygiene.
For web links, fetch the full visible article text. If the initial extractor returns a summary or truncation, pull the page HTML directly and strip scripts/styles before judging.
Scoring Modes
Classification Mode (used by editor-in-chief)
When invoked by the editor-in-chief skill, use classification labels instead of numeric scores. This maps the 6 questions to 6 dimensions with STRONG / NEEDS WORK / WEAK labels:
| Question | Dimension | STRONG | NEEDS WORK | WEAK |
|---|
| Q1 (Shareable?) | Shareability | 2+ screenshot moments, reader would forward | Hook/insight exists but buried or undersold | Nothing worth sharing |
| Q2 (Informative/fun?) | Substance | Every claim backed by evidence | Mix of evidence and assertions | Vague claims, no proof |
| Q3 (Human voice?) | Voice | Sounds like the author, irregular rhythm, opinionated | Mostly human, some stiff patches or AI tells | Robot cadence, banned patterns present |
| Q4 (Cuttable fat?) | Leanness | Every sentence earns its place | 10-20% filler | 30%+ chaff |
| Q5 (Unique angle?) | Emotion | Driving emotion clear and felt throughout | Emotion exists but buried or inconsistent | Flat, no emotional throughline |
| Q6 (Target reader?) | Reader Fit | Clear reader profile, article directly serves them | General audience, not specific enough | Nobody in particular |
Note: editor-in-chief uses the first five rows plus its own Rhythm dimension, and defers Reader Fit to its final reader-simulator gate.
Output format for classification mode:
Shareability: [STRONG|NEEDS WORK|WEAK]: [1-2 sentence explanation with specific examples]
Substance: [STRONG|NEEDS WORK|WEAK]: [explanation]
Voice: [STRONG|NEEDS WORK|WEAK]: [explanation]
Leanness: [STRONG|NEEDS WORK|WEAK]: [explanation]
Emotion: [STRONG|NEEDS WORK|WEAK]: [explanation]
Reader Fit: [STRONG|NEEDS WORK|WEAK]: [explanation]
Numeric Mode (standalone use)
When used standalone (not via editor-in-chief), use the original numeric scoring:
| Score | Meaning |
|---|
| 28-30 | Ship it. This is great. |
| 22-27 | Good but needs polish. Address the weakest scores. |
| 16-21 | Needs significant rework. Focus on scores below 3. |
| Below 16 | Start over or fundamentally rethink the angle. |
Output Format (Numeric Mode)
Match report length to the input. For a headline or a tweet, a three-line verdict is the whole deliverable. Drop sections that would restate the score table.
Your own feedback obeys the same rules: no fake insider framing, no formulaic contrast openers, no throat-clearing, no vague rewrites.
# Content Evaluation: [title or description]
## Scores
| Question | Score | One-line verdict |
|----------|-------|-----------------|
| Shareable? | X/5 | ... |
| Informative/fun? | X/5 | ... |
| Human voice? | X/5 | ... |
| Cuttable fat? | X/5 | ... |
| Unique angle? | X/5 | ... |
| Target reader? | X/5 | ... |
| **Total** | **XX/30** | |
## Detailed Feedback
### Shareable? (X/5)
[explanation + specific examples from the content]
### Informative/fun? (X/5)
[explanation + specific examples]
### Human voice? (X/5)
[explanation + flagged slop patterns]
### Cuttable fat? (X/5)
[specific cut list]
### Unique angle? (X/5)
[thesis + driving emotion + "only I" factor]
### Target reader? (X/5)
[reader profile + trigger + search query + awareness level]
## Top 3 Changes That Would Improve This Most
1. ...
2. ...
3. ...
Division of labour with the writer skill
This skill scores. writer fixes and owns voice.
- Scoring rubric (the six dimensions, STRONG/NEEDS WORK/WEAK): here.
- Voice ground truth (kill phrases, fingerprints):
writer/references/WRITING-STYLE.md.
- The 24 AI tells and their fixes:
writer/references/humanizer-checklist.md.
- Structural tells and thresholds:
writer/references/structural-tells.md.
When scoring Voice, read those files rather than judging on generic "sounds human". Do not
copy their contents back into this skill.
Hard Filters
Before scoring, check these. Any hit is a red flag that should pull the relevant score down:
- Redundancy: If any sentence merely restates the previous one in different words, flag it. One thought, one sentence.
- Emotion-guiding: If any sentence exists only to tell the reader how to feel ("This is exciting", "What's remarkable is"), flag it. The content should create the feeling, not name it.
- Uncertain padding: If quality is uncertain on a section, the fix is cutting, not adding. Less confident = write less. Silence beats slop.
- Fake insider framing: Flag any sentence using "The part nobody talks about..." / "What they don't tell you..." / "The real secret is..." / "Most people miss this..." / "Here's what most people get wrong..." (these imply secret knowledge without delivering anything genuinely exclusive). Pull Voice score down.
- Em dashes: Flag every em dash. The rule is no em dashes, ever. Each one is a Voice deduction; the fix is a period, comma, colon, semicolon, or parentheses.
Do not flag these
The filters above will damage good writing if applied to everything. These are signs of a
human writing well. Leave them:
- A long sentence that runs long because the thought is genuinely tangled
- An odd but precise word
- An aside that isn't load-bearing
- Uneven section lengths
- The same noun repeated instead of a synonym
- A flat, unadorned statement with no rhetorical shape
- A blunt, unhedged opinion
- Detail that reads as excessive specificity
Two more limits when you edit rather than score:
- Rewrite as much as the draft needs. Just don't add a number, source, or personality
that wasn't already there. A made-up specific is worse than the vague line it replaced,
because vague can be fixed and wrong can't. Missing the real detail, flag the gap.
(This is about cleanup. If the user asked you to write or expand something, write it.)
- Don't over-edit. Apply every rule at full strength and you get the sameness you were
trying to remove. When a rule fights the author's voice, the voice wins.
Short-Form Video Hook Checklist
Use this when evaluating TikTok slideshows, YouTube Shorts scripts, or any short-form video content. Run before the 6 questions above.
A solo dev with zero audience, zero budget got 18M views in 28 days using these principles. Each is a binary pass/fail.
Hook: Is it about the viewer, not the product?
- ❌ "Bloom's AI analyzes stocks for you"
- ✅ "I let an AI pick my stocks for 30 days, here's what it found"
If the first line is about the product's features, it will flop. Rewrite until it opens with something the viewer experiences, fears, wants, or wonders about themselves.
Angle: Is it broad enough for people outside your niche?
- ❌ "Here's what this earnings report means for $NVDA investors"
- ✅ "Everyone laughed when I bought this stock. Now they're asking me how."
Can a non-investor relate to it? Can a non-user of your app still feel something from the opening? If only your target customer would watch past 3 seconds, the angle is too narrow.
Curiosity gap: Is there something they need to see by the end?
The hook must create a question the viewer needs answered. Tease a result, reveal, or test early, hold the answer until the final slide. If there's no unresolved tension, there's nothing keeping them to the end.
15-second silent demo test: Does it work with sound off?
Watch the first 15 seconds muted. If a stranger couldn't understand what the content is about or feel the hook's tension without audio, the visual hook is broken, and no amount of copy fixes a broken visual hook. Captions help but don't substitute for a self-explanatory visual sequence.
Line discipline: Does every line earn the next?
Read the script line by line. For each line, ask: does this build curiosity, escalate tension, or deliver on a promise? If it explains, teaches, or provides context without earning it, cut it. Information is the reward at the end, not the scaffolding throughout.
The Seven Sweeps Framework (Copy Editing)
Edit copy through seven sequential passes, each focusing on one dimension. The sweeps are lenses, not gates. Apply them in one pass and keep earlier lenses in mind as you edit.
Sweep 1: Clarity
Focus: Can the reader understand what you're saying?
What to check:
- Confusing sentence structures
- Unclear pronoun references
- Jargon or insider language
- Ambiguous statements
- Missing context
Common issues:
- Sentences trying to say too much
- Abstract language instead of concrete
- Assuming reader knowledge they don't have
- Burying the point in qualifications
Process:
- Read through quickly, highlighting unclear parts
- Don't correct yet: just note problem areas
- After marking issues, recommend specific edits
- Confirm the "Rule of One" (one main idea per section) and "You Rule" (copy speaks to the reader) are intact
Sweep 2: Voice and Tone
Focus: Is the copy consistent in how it sounds?
What to check:
- Shifts between formal and casual
- Inconsistent brand personality
- Mood changes that feel jarring
- Word choices that don't match the brand
Common issues:
- Starting casual, becoming corporate
- Mixing "we" and "the company" references
- Humor in some places, serious in others (unintentionally)
- Technical language appearing randomly
Process:
- Read aloud to hear inconsistencies
- Mark where tone shifts unexpectedly
- Recommend edits that smooth transitions
- Ensure personality remains throughout
Sweep 3: So What
Focus: Does every claim answer "why should I care?"
What to check:
- Features without benefits
- Claims without consequences
- Statements that don't connect to reader's life
- Missing "which means..." bridges
The So What test: For every statement, ask "Okay, so what?" If the copy doesn't answer with a deeper benefit, it needs work.
- ❌ "Our platform uses AI-powered analytics"
- ✅ "Our analytics surface what you'd miss reading the raw data, so a decision takes an hour instead of a day"
Process:
- Read each claim and literally ask "so what?"
- Highlight claims missing the answer
- Add the benefit bridge or deeper meaning
- Ensure benefits connect to real reader desires
Sweep 4: Prove It
Focus: Is every claim supported with evidence?
What to check:
- Unsubstantiated claims
- Missing social proof
- Assertions without backup
- "Best" or "leading" without evidence
Types of proof: Testimonials with names, case study references, statistics and data, third-party validation, guarantees and risk reversals, customer logos, review scores.
Common issues:
- "Trusted by thousands" (which thousands?)
- "Industry-leading" (according to whom?)
- "Customers love us" (show them saying it)
Process:
- Identify every claim that needs proof
- Check if proof exists nearby
- Flag unsupported assertions
- Recommend adding proof or softening claims
Sweep 5: Specificity
Focus: Is the copy concrete enough to be compelling?
What to check:
- Vague language ("improve," "enhance," "optimize")
- Generic statements that could apply to anyone
- Round numbers that feel made up
- Missing details that would make it real
Specificity upgrades:
| Vague | Specific |
|---|
| Save time | Save 4 hours every week |
| Many customers | 2,847 teams |
| Fast results | Results in 14 days |
| Improve your workflow | Cut your reporting time in half |
| Great support | Response within 2 hours |
Process:
- Highlight vague words and phrases
- Ask "Can this be more specific?"
- Add numbers, timeframes, or examples
- Remove content that can't be made specific (it's probably filler)
Sweep 6: Heightened Emotion
Focus: Does the copy make the reader feel something?
What to check:
- Flat, informational language
- Missing emotional triggers
- Pain points mentioned but not felt
- Aspirations stated but not evoked
Techniques: Paint the "before" state vividly, use sensory language, tell micro-stories, reference shared experiences, ask questions that prompt reflection.
Process:
- Read for emotional impact: does it move you?
- Identify flat sections that should resonate
- Add emotional texture while staying authentic
- Ensure emotion serves the message (not manipulation)
Sweep 7: Zero Risk
Focus: Have we removed every barrier to action?
What to check:
- Friction near CTAs
- Unanswered objections
- Missing trust signals
- Unclear next steps
- Hidden costs or surprises
Risk reducers: Money-back guarantees, free trials, "No credit card required," "Cancel anytime," social proof near CTA, clear expectations of what happens next, privacy assurances.
Process:
- Focus on sections near CTAs
- List every reason someone might hesitate
- Check if the copy addresses each concern
- Add risk reversals or trust signals as needed
Quick-Pass Editing Checks
Use these for faster reviews when a full seven-sweep process isn't needed.
Word-Level Checks
Cut these words:
- Very, really, extremely, incredibly (weak intensifiers)
- Just, actually, basically (filler)
- In order to (use "to")
- That (often unnecessary)
- Things, stuff (vague)
Replace these:
| Weak | Strong |
|---|
| Utilize | Use |
| Implement | Set up |
| Leverage | Use |
| Facilitate | Help |
| Innovative | New |
| Robust | Strong |
| Seamless | Smooth |
| Cutting-edge | New/Modern |
Plain English Alternatives
Replace complex or pompous words with simpler ones. Source: Plain English Campaign, plainlanguage.gov.
| Complex | Plain |
|---|
| absence of | no, none |
| accomplish | do, finish |
| additional | extra, more |
| advise | tell, say |
| allocate | give, share |
| anticipate | expect |
| approximately | about |
| ascertain | find out |
| assistance | help |
| at the present time | now |
| cease | stop, end |
| commence | start, begin |
| communicate | tell, talk |
| consequently | so |
| currently | now |
| demonstrate | show, prove |
| determine | decide |
| discontinue | stop |
| disseminate | spread |
| due to the fact that | because |
| endeavour | try |
| establish | set up, show |
| expedite | speed up |
| facilitate | help |
| for the purpose of | to, for |
| furthermore | also, and |
| implement | carry out, do |
| in accordance with | under |
| in conjunction with | with |
| in order to | to |
| in the event of | if |
| indicate | show, suggest |
| initiate | start, begin |
| moreover | also, and |
| notify | tell |
| obtain | get |
| on behalf of | for |
| owing to | because |
| permit | let, allow |
| prior to | before |
|
Phrases to remove entirely: "a total of," "absolutely," "actually," "at the end of the day," "at this moment in time," "basically," "I am of the opinion that" (use "I think"), "in the final analysis," "it should be understood," "last but not least," "obviously," "of course," "quite," "really," "the fact of the matter is," "to all intents and purposes," "very."
Watch for:
- Adverbs (usually unnecessary)
- Passive voice (switch to active)
- Nominalizations (verb turned noun: "make a decision" → "decide")
Sentence-Level Checks
These are defaults for corporate and marketing copy, not laws. A sentence that breaks
them because the thought genuinely runs that long is on the "Do not flag these" list
above. Check that list before flagging any of these.
- One idea per sentence
- Vary sentence length (mix short and long)
- Front-load important information
- Max 3 conjunctions per sentence
- No more than 25 words (usually)
Paragraph-Level Checks
- One topic per paragraph
- Short paragraphs (2-4 sentences for web)
- Strong opening sentences
- Logical flow between paragraphs
- White space for scannability
Common Copy Problems & Fixes
| Problem | Symptom | Fix |
|---|
| Wall of Features | List of what the product does without why it matters | Add "which means..." after each feature to bridge to benefits |
| Corporate Speak | "Leverage synergies to optimize outcomes" | Ask "How would a human say this?" and use those words |
| Weak Opening | Starting with company history or vague statements | Lead with the reader's problem or desired outcome |
| Buried CTA | The ask comes after too much buildup, or isn't clear | Make the CTA obvious, early, and repeated |
| No Proof | "Customers love us" with no evidence | Add specific testimonials, numbers, or case references |
| Generic Claims | "We help businesses grow" | Specify who, how, and by how much |
| Mixed Audiences | Copy tries to speak to everyone, resonates with no one | Pick one audience and write directly to them |
| Feature Overload | Listing every capability, overwhelming the reader | Focus on 3-5 key benefits that matter most to the audience |
Load on-demand: references/pillar-evaluation.md for monthly pillar-level evaluation: scorecard, verdicts (SCALE/KEEP/ELEVATE/ROTATE OUT), recap format, and pillar-vs-post-level comparison.
When to Use This
- Before publishing any article: run the six questions as final gate
- On thread hooks: just questions 1, 3, and 5
- On tweet batches: just questions 1 and 3
- On headlines: just questions 1 and 5
- When something underperforms: diagnose why with all 6
- Monthly content review: run pillar-level evaluation across all active pillars
References
- Used by article-writer (revision pass)
- Used by typefully (hook quality check)
- Used by tweet-ideas (tweet quality)
- Used by hooks (title evaluation)
- Shares voice standards with article-writer humanizer section
- the
writer skill's references/WRITING-STYLE.md: ground truth for voice evaluation. Read the Kill Phrases list and Voice Fingerprint section before scoring Question 3. Generic "sounds human" is not the bar; Eric's specific fingerprint is.
