Skill: humanistic_reviewer
Purpose
Audit draft copy for AI slop, generic filler, voice mismatch, and low signal. Provide targeted feedback, keep the strongest parts, and optionally rewrite.
Inputs
| Field |
Type |
Default |
Description |
draft_copy |
text |
required |
Copy to review |
asset_type |
enum |
required |
email, webinar, funnel_page, website, ad |
brand_voice |
string[] |
["direct", "no-BS", "punchy", "mentor-not-guru"] |
2–4 adjectives |
audience |
string |
required |
Short description of who this is for |
max_length |
int |
optional |
Desired length ceiling |
strictness_level |
int 1–3 |
2 |
1=lenient, 2=normal, 3=ruthless |
rewrite_threshold |
int 0–10 |
6 |
Score below this triggers auto-rewrite |
auto_rewrite |
boolean |
false |
If true, produce revised_draft in output |
Outputs
| Field |
Type |
Description |
overall_score |
int 0–10 |
Quality score |
slop_flags |
array |
Issues found (see structure below) |
revision_instructions |
string[] |
3–7 actionable bullets |
keep_sections |
string[] |
Strong excerpts to preserve |
rewrite_needed |
boolean |
True if score < rewrite_threshold |
revised_draft |
text |
Only present if auto_rewrite=true or rewrite_needed=true |
slop_flags structure
{
"type": "generic | repetitive | bloated | robotic | off_brand | unclear | cliche | weak_hook | buried_cta",
"excerpt": "exact text from draft",
"issue_explanation": "why this is a problem",
"suggested_fix": "specific replacement or direction"
}
Behavior Rules
1. Kill Generic Filler
Flag and require replacement for:
- "in today's fast-paced world"
- "ever-changing landscape"
- "leverage the power of"
- "unlock your potential"
- "game-changing" / "groundbreaking" / "revolutionary"
- "journey" (when used metaphorically)
- "dive deep" / "deep dive"
- "at the end of the day"
- "it's no secret that"
- "more than ever before"
Rule: Replace with specific, concrete language. No exceptions.
2. Kill Over-Explaining
- Flag intros longer than 3 sentences before any value is delivered.
- Flag the same idea stated more than once in different words.
- Instruct cuts, not rewrites of bloat.
3. Enforce Clear Thinking
- Ask: "What is the single core idea?"
- If the copy tries to say 3 things, flag it: "Pick one. What does the reader need to believe to take action?"
- Require that the core idea appears in the first 3 sentences.
4. Enforce Voice Alignment
Compare draft tone against brand_voice. Flag mismatches:
- Too formal vs "direct/no-BS" → flag as
off_brand
- Too casual vs "premium" → flag as
off_brand
- Guru-speak vs "mentor-not-guru" → flag as
off_brand
5. Enforce Asset-Specific Fit
Ads:
- Hook must be in the first sentence. If not →
weak_hook.
- One idea only. Multiple angles in one ad →
bloated.
- CTA must be explicit. Vague ending →
buried_cta.
Emails:
- Subject line must create curiosity or state a clear benefit. No puns that don't land.
- One primary CTA. Two CTAs → flag as
bloated.
- Skimmable: long unbroken paragraphs (>4 lines) → flag as
bloated.
Webinar copy:
- Big idea must be visible in the headline. Generic outcome →
unclear.
- Proof must be specific. Vague testimonials →
generic.
- Bullet list features without outcomes → flag as
generic.
Sales / funnel pages:
- Above-fold must answer: What is this? Who is it for? Why now?
- Missing any of three →
unclear.
- Guarantee must be specific. "100% satisfaction" alone is weak → flag.
Scoring Guide
| Score |
Meaning |
| 9–10 |
Ship it. Strong hook, clear idea, voice-matched, proof-specific. |
| 7–8 |
Good. Minor polish needed. No structural rewrites. |
| 5–6 |
Weak. Structural issues. Requires meaningful revision before shipping. |
| 3–4 |
Not ready. Core idea is unclear or voice is off. Full rewrite recommended. |
| 0–2 |
Start over. Generic, hallucinated, or completely off-brand. |
A score below rewrite_threshold (default: 6) sets rewrite_needed: true.
If auto_rewrite: true or rewrite_needed: true, produce a revised_draft that:
- Applies all revision instructions
- Preserves all
keep_sections
- Does not invent new proof; keeps placeholders
- Matches
brand_voice
Example Revision Instructions (Pattern)
- "Cut 20–30% by removing repeated points and vague phrases."
- "Move the core promise into the first 2 sentences."
- "Replace '[generic adjective]' with a specific outcome — time saved, revenue added, or pain removed."
- "Sharpen the CTA to one clear action. Remove the secondary link."
- "The subject line is clever but unclear. Swap for benefit-forward: tell them what's inside."
- "The opening paragraph is preamble. Delete it. Start at line 3."
- "Replace '[testimonial placeholder]' with a real name, result, and timeframe."
1---2name: humanistic-reviewer3description: Audit any draft copy for AI slop, generic filler, voice mismatch, and weak structure. Use this skill after any writing engine produces output — always run before delivering copy to the user. Also trigger when the user says "review this copy", "edit this", "make this less AI-sounding", "punch this up", "is this good enough to send", or pastes copy and asks for feedback. Returns a score, flagged issues, revision instructions, and optionally a rewritten draft.4---56# Skill: humanistic_reviewer78## Purpose9Audit draft copy for AI slop, generic filler, voice mismatch, and low signal. Provide targeted feedback, keep the strongest parts, and optionally rewrite.1011---1213## Inputs1415| Field | Type | Default | Description |16|---|---|---|---|17| `draft_copy` | text | required | Copy to review |18| `asset_type` | enum | required | `email`, `webinar`, `funnel_page`, `website`, `ad` |19| `brand_voice` | string[] | `["direct", "no-BS", "punchy", "mentor-not-guru"]` | 2–4 adjectives |20| `audience` | string | required | Short description of who this is for |21| `max_length` | int | optional | Desired length ceiling |22| `strictness_level` | int 1–3 | `2` | 1=lenient, 2=normal, 3=ruthless |23| `rewrite_threshold` | int 0–10 | `6` | Score below this triggers auto-rewrite |24| `auto_rewrite` | boolean | `false` | If true, produce revised_draft in output |2526---2728## Outputs2930| Field | Type | Description |31|---|---|---|32| `overall_score` | int 0–10 | Quality score |33| `slop_flags` | array | Issues found (see structure below) |34| `revision_instructions` | string[] | 3–7 actionable bullets |35| `keep_sections` | string[] | Strong excerpts to preserve |36| `rewrite_needed` | boolean | True if score < rewrite_threshold |37| `revised_draft` | text | Only present if auto_rewrite=true or rewrite_needed=true |3839### slop_flags structure40```json41{42 "type": "generic | repetitive | bloated | robotic | off_brand | unclear | cliche | weak_hook | buried_cta",43 "excerpt": "exact text from draft",44 "issue_explanation": "why this is a problem",45 "suggested_fix": "specific replacement or direction"46}47```4849---5051## Behavior Rules5253### 1. Kill Generic Filler54Flag and require replacement for:55- "in today's fast-paced world"56- "ever-changing landscape"57- "leverage the power of"58- "unlock your potential"59- "game-changing" / "groundbreaking" / "revolutionary"60- "journey" (when used metaphorically)61- "dive deep" / "deep dive"62- "at the end of the day"63- "it's no secret that"64- "more than ever before"6566Rule: Replace with specific, concrete language. No exceptions.6768### 2. Kill Over-Explaining69- Flag intros longer than 3 sentences before any value is delivered.70- Flag the same idea stated more than once in different words.71- Instruct cuts, not rewrites of bloat.7273### 3. Enforce Clear Thinking74- Ask: "What is the single core idea?"75- If the copy tries to say 3 things, flag it: "Pick one. What does the reader need to believe to take action?"76- Require that the core idea appears in the first 3 sentences.7778### 4. Enforce Voice Alignment79Compare draft tone against `brand_voice`. Flag mismatches:80- Too formal vs "direct/no-BS" → flag as `off_brand`81- Too casual vs "premium" → flag as `off_brand`82- Guru-speak vs "mentor-not-guru" → flag as `off_brand`8384### 5. Enforce Asset-Specific Fit8586**Ads:**87- Hook must be in the first sentence. If not → `weak_hook`.88- One idea only. Multiple angles in one ad → `bloated`.89- CTA must be explicit. Vague ending → `buried_cta`.9091**Emails:**92- Subject line must create curiosity or state a clear benefit. No puns that don't land.93- One primary CTA. Two CTAs → flag as `bloated`.94- Skimmable: long unbroken paragraphs (>4 lines) → flag as `bloated`.9596**Webinar copy:**97- Big idea must be visible in the headline. Generic outcome → `unclear`.98- Proof must be specific. Vague testimonials → `generic`.99- Bullet list features without outcomes → flag as `generic`.100101**Sales / funnel pages:**102- Above-fold must answer: What is this? Who is it for? Why now?103- Missing any of three → `unclear`.104- Guarantee must be specific. "100% satisfaction" alone is weak → flag.105106---107108## Scoring Guide109110| Score | Meaning |111|---|---|112| 9–10 | Ship it. Strong hook, clear idea, voice-matched, proof-specific. |113| 7–8 | Good. Minor polish needed. No structural rewrites. |114| 5–6 | Weak. Structural issues. Requires meaningful revision before shipping. |115| 3–4 | Not ready. Core idea is unclear or voice is off. Full rewrite recommended. |116| 0–2 | Start over. Generic, hallucinated, or completely off-brand. |117118A score below `rewrite_threshold` (default: 6) sets `rewrite_needed: true`.119120If `auto_rewrite: true` or `rewrite_needed: true`, produce a `revised_draft` that:121- Applies all revision instructions122- Preserves all `keep_sections`123- Does not invent new proof; keeps placeholders124- Matches `brand_voice`125126---127128## Example Revision Instructions (Pattern)129- "Cut 20–30% by removing repeated points and vague phrases."130- "Move the core promise into the first 2 sentences."131- "Replace '[generic adjective]' with a specific outcome — time saved, revenue added, or pain removed."132- "Sharpen the CTA to one clear action. Remove the secondary link."133- "The subject line is clever but unclear. Swap for benefit-forward: tell them what's inside."134- "The opening paragraph is preamble. Delete it. Start at line 3."135- "Replace '[testimonial placeholder]' with a real name, result, and timeframe."
automatedmarketer