Brand Review (Content)
Quality gate between drafting and posting. Optimized for indie-hacker / build-in-public content where authenticity is the moat. Adapted from Anthropic's marketing/brand-review skill, stripped of corporate-compliance framing.
When to Use
- User pastes a draft and asks "is this good", "review this", "any feedback"
- Output of
draft-post needs a second pass before shipping
- User wants to audit existing posts to catch voice drift over time
When NOT to Use
- Initial drafting → use
draft-post
- Reply to someone else's post → use
reply-research
- Documenting brand voice from scratch → see the "Brand Voice Documentation" section below, then save the result somewhere persistent (skill, memory, or vault note)
Inputs
- Content to review — paste, file path, or URL. Accept multiple pieces for batch review.
- Channel context — X, LinkedIn, blog, etc. Standards shift by channel.
- Brand voice source — if user has documented voice (in this skill's reference files, memory, or vault), apply it. Otherwise use the defaults below and offer to document afterward.
Review Dimensions
1. AI-ism Detection (the highest-severity check for indie content)
Flag any of these as HIGH severity:
| Pattern |
Why it kills authenticity |
| "leverage", "unlock", "game-changer", "revolutionize" |
Universal corporate-AI tells |
| "seamless", "holistic", "robust", "synergy", "cutting-edge" |
Same |
| "In today's fast-paced world..." |
Opener used by every AI blog |
| "Let's dive in" / "Let's unpack" |
Pure filler |
| "It's important to note that..." |
Hedge that adds nothing |
| "This is a game-changer for X" |
Manufactured stakes |
| Three-bullet structures where one sentence works |
Listicle bloat |
| Em-dashes followed by reformulations of the same point |
Padding |
| "Not just X, but Y" parallel structure repeated |
AI rhythm tell |
| Tweets that end with engagement bait ("Thoughts?", "Agree?") |
Generic |
Flag and propose a rewrite for every instance.
2. Specificity Check
For build-in-public content specifically:
- Are there real numbers? ("47 users" beats "some users")
- Are products/tools named? ("Hermes on a Hetzner CX22" beats "my server")
- Are failures named? Or only wins? (Failure stories drive 2-3x engagement)
- Is there a real trigger event? (What happened today, not "thoughts on X")
Flag vague claims as MEDIUM. Push the writer toward specifics.
3. Voice Consistency
If brand voice is documented (see Documentation Framework below), compare each section against:
- Defined voice attributes (e.g. "warm, dry, technical")
- Tone calibration for the channel (X is punchier than LinkedIn)
- Preferred vs avoided terms
- POV consistency (don't drift between "I" and "we")
If voice is undocumented, evaluate against generic indie-hacker defaults:
- First-person consistent
- Past tense for stories, present tense for live updates
- No marketing-speak
- Honest about scope and impact
4. Structural Hygiene
- Hook — does the first 8 words earn the click?
- One idea — single tweet should have one point, not three
- Threads — does each tweet stand alone? Is the structure (setup → body → reversal → CTA) intact?
- CTAs — only present if real. No manufactured "DM me" lines.
- Length — tight is better. Cut filler ruthlessly.
5. Authenticity Flags
Always check:
- Manufactured stakes — "This changed everything" when it didn't
- Borrowed authority — claims that imply expertise the writer doesn't have
- Fake humility — "I'm just a humble builder" patterns
- Inflated metrics — rounding, vanity numbers, cherry-picked timeframes
- Stolen frames — phrasing that's clearly lifted from another big account
6. Hard No's (rarely needed but worth checking)
- Unsubstantiated comparative claims about real competitors
- Health/financial/legal claims without grounding
- Quotes attributed without permission
- Content that punches down
Output Format
## Brand Review: [Content type]
### Summary
- **Verdict:** [Ship / Ship after small fixes / Rework needed]
- **Strengths (1-2 lines):** [what's working]
- **Main issues (1-2 lines):** [headline problems]
### Findings
| # | Issue | Location | Severity | Fix |
|---|-------|----------|----------|-----|
| 1 | [issue] | [quote or section] | High/Med/Low | [specific rewrite] |
### Revised Sections (top 3-5 high-severity)
**Before:**
> [original quote]
**After:**
> [proposed rewrite]
**Why:** [1 line]
### Open Questions
[anything you need from the user to refine further]
After Review
Offer one clear next step. Don't list a menu:
- If verdict is "Ship after small fixes" → "Want me to apply the highlighted fixes and give you the final version?"
- If verdict is "Rework needed" → "The hook needs to change before anything else lands. Want me to redraft from the hook down?"
- If verdict is "Ship" → "Looks good. Want me to post via xurl or hand it back to you?"
Brand Voice Documentation Framework
When the user wants to define their voice (often after a few reviews), use the template at references/brand-voice-template.md. It has pre-built tables for each section — the user fills it in, the skill references it on every review.
Minimum useful structure (see template for full detail):
- Personality — describe the brand as a person (1 paragraph)
- Voice attributes — 3-5 attributes, each with:
- "We are: [what this means in practice]"
- "We are not: [common misinterpretation]"
- "This sounds like: [example]"
- "This does NOT sound like: [example]"
- Tone calibration — how voice shifts across X / LinkedIn / blog / email
- Preferred terms / avoided terms — concrete list
- POV rules — first person? "I" or "we"?
Voice Attribute Spectrums (for picking attributes)
| Spectrum |
One End |
Other End |
| Formality |
Formal, institutional |
Casual, conversational |
| Authority |
Expert, authoritative |
Peer-level, collaborative |
| Emotion |
Warm, empathetic |
Direct, matter-of-fact |
| Complexity |
Technical, precise |
Simple, accessible |
| Energy |
Bold, energetic |
Calm, measured |
| Humor |
Playful, witty |
Serious, earnest |
Indie-hacker defaults usually land: casual, peer-level, direct, technical-but-accessible, calm, dry-humor.
Pitfalls
Don't be precious about the user's draft. If something is mid, say so. Flattery hurts the writer.
Don't auto-rewrite without permission. Flag issues, propose fixes, let the user decide. Their voice > your suggestion.
Don't over-flag on Low-severity issues. If a post is 95% good, lead with that and surface only the High/Medium issues. Drowning the user in nitpicks is its own failure mode.
Calibrate severity to channel. A "leverage" in a LinkedIn post is Medium. In a build-in-public tweet, it's High — different audiences tolerate different things.
Don't review your own drafts. If draft-post produced the content in the same session, the writer's-block bias applies. Ask the user to read it first OR explicitly note "self-review, treat skeptically."
Voice drift is real. If the user has documented voice, compare to it directly. If they've been posting for 6+ months, the voice has probably evolved — offer to re-document after 5+ reviews.
Related Skills
draft-post — drafting layer (the thing this skill reviews)
build-in-public — strategy layer (cadence, X Premium, infra)
reply-research — replies are a different beast; this skill is for original content
1---
2name: brand-review-content
3description: Review a content draft (X post, thread, LinkedIn post, blog, email) against brand voice and authenticity standards before it ships. Flags AI-isms, corporate jargon, unsubstantiated claims, manufactured stakes, and voice inconsistencies. Provides severity-rated findings with before/after fixes. Use after drafting and before posting.
4version: 1.0.0
5author: senna
6triggers:
7 - "review this draft"
8 - "check this post"
9 - "is this on-brand"
10 - "audit this content"
11 - "brand review"
12metadata:
13 source: https://skillmd.com/skills/theheavenlyd3mon/brand-review-content
14 hermes:
15 tags: [social-media, content, review, qa, brand-voice, authenticity]
16 related_skills: [draft-post, build-in-public, reply-research]
17---
18
19# Brand Review (Content)
20
21Quality gate between drafting and posting. Optimized for indie-hacker / build-in-public content where authenticity is the moat. Adapted from Anthropic's marketing/brand-review skill, stripped of corporate-compliance framing.
22
23## When to Use
24
25- User pastes a draft and asks "is this good", "review this", "any feedback"
26- Output of `draft-post` needs a second pass before shipping
27- User wants to audit existing posts to catch voice drift over time
28
29## When NOT to Use
30
31- Initial drafting → use `draft-post`
32- Reply to someone else's post → use `reply-research`
33- Documenting brand voice from scratch → see the "Brand Voice Documentation" section below, then save the result somewhere persistent (skill, memory, or vault note)
34
35## Inputs
36
371. **Content to review** — paste, file path, or URL. Accept multiple pieces for batch review.
382. **Channel context** — X, LinkedIn, blog, etc. Standards shift by channel.
393. **Brand voice source** — if user has documented voice (in this skill's reference files, memory, or vault), apply it. Otherwise use the defaults below and offer to document afterward.
40
41## Review Dimensions
42
43### 1. AI-ism Detection (the highest-severity check for indie content)
44
45Flag any of these as **HIGH severity**:
46
47| Pattern | Why it kills authenticity |
48|---------|---------------------------|
49| "leverage", "unlock", "game-changer", "revolutionize" | Universal corporate-AI tells |
50| "seamless", "holistic", "robust", "synergy", "cutting-edge" | Same |
51| "In today's fast-paced world..." | Opener used by every AI blog |
52| "Let's dive in" / "Let's unpack" | Pure filler |
53| "It's important to note that..." | Hedge that adds nothing |
54| "This is a game-changer for X" | Manufactured stakes |
55| Three-bullet structures where one sentence works | Listicle bloat |
56| Em-dashes followed by reformulations of the same point | Padding |
57| "Not just X, but Y" parallel structure repeated | AI rhythm tell |
58| Tweets that end with engagement bait ("Thoughts?", "Agree?") | Generic |
59
60Flag and propose a rewrite for every instance.
61
62### 2. Specificity Check
63
64For build-in-public content specifically:
65
66- Are there real numbers? ("47 users" beats "some users")
67- Are products/tools named? ("Hermes on a Hetzner CX22" beats "my server")
68- Are failures named? Or only wins? (Failure stories drive 2-3x engagement)
69- Is there a real trigger event? (What happened today, not "thoughts on X")
70
71Flag vague claims as **MEDIUM**. Push the writer toward specifics.
72
73### 3. Voice Consistency
74
75If brand voice is documented (see Documentation Framework below), compare each section against:
76
77- Defined voice attributes (e.g. "warm, dry, technical")
78- Tone calibration for the channel (X is punchier than LinkedIn)
79- Preferred vs avoided terms
80- POV consistency (don't drift between "I" and "we")
81
82If voice is undocumented, evaluate against generic indie-hacker defaults:
83
84- First-person consistent
85- Past tense for stories, present tense for live updates
86- No marketing-speak
87- Honest about scope and impact
88
89### 4. Structural Hygiene
90
91- **Hook** — does the first 8 words earn the click?
92- **One idea** — single tweet should have one point, not three
93- **Threads** — does each tweet stand alone? Is the structure (setup → body → reversal → CTA) intact?
94- **CTAs** — only present if real. No manufactured "DM me" lines.
95- **Length** — tight is better. Cut filler ruthlessly.
96
97### 5. Authenticity Flags
98
99Always check:
100
101- **Manufactured stakes** — "This changed everything" when it didn't
102- **Borrowed authority** — claims that imply expertise the writer doesn't have
103- **Fake humility** — "I'm just a humble builder" patterns
104- **Inflated metrics** — rounding, vanity numbers, cherry-picked timeframes
105- **Stolen frames** — phrasing that's clearly lifted from another big account
106
107### 6. Hard No's (rarely needed but worth checking)
108
109- Unsubstantiated comparative claims about real competitors
110- Health/financial/legal claims without grounding
111- Quotes attributed without permission
112- Content that punches down
113
114## Output Format
115
116```
117## Brand Review: [Content type]
118
119### Summary
120- **Verdict:** [Ship / Ship after small fixes / Rework needed]
121- **Strengths (1-2 lines):** [what's working]
122- **Main issues (1-2 lines):** [headline problems]
123
124### Findings
125
126| # | Issue | Location | Severity | Fix |
127|---|-------|----------|----------|-----|
128| 1 | [issue] | [quote or section] | High/Med/Low | [specific rewrite] |
129
130### Revised Sections (top 3-5 high-severity)
131
132**Before:**
133> [original quote]
134
135**After:**
136> [proposed rewrite]
137
138**Why:** [1 line]
139
140### Open Questions
141[anything you need from the user to refine further]
142```
143
144## After Review
145
146Offer one clear next step. Don't list a menu:
147
148- If verdict is "Ship after small fixes" → "Want me to apply the highlighted fixes and give you the final version?"
149- If verdict is "Rework needed" → "The hook needs to change before anything else lands. Want me to redraft from the hook down?"
150- If verdict is "Ship" → "Looks good. Want me to post via xurl or hand it back to you?"
151
152## Brand Voice Documentation Framework
153
154When the user wants to define their voice (often after a few reviews), use the template at [`references/brand-voice-template.md`](references/brand-voice-template.md). It has pre-built tables for each section — the user fills it in, the skill references it on every review.
155
156Minimum useful structure (see template for full detail):
157
1581. **Personality** — describe the brand as a person (1 paragraph)
1592. **Voice attributes** — 3-5 attributes, each with:
160 - "We are: [what this means in practice]"
161 - "We are not: [common misinterpretation]"
162 - "This sounds like: [example]"
163 - "This does NOT sound like: [example]"
1643. **Tone calibration** — how voice shifts across X / LinkedIn / blog / email
1654. **Preferred terms / avoided terms** — concrete list
1665. **POV rules** — first person? "I" or "we"?
167
168### Voice Attribute Spectrums (for picking attributes)
169
170| Spectrum | One End | Other End |
171|----------|---------|-----------|
172| Formality | Formal, institutional | Casual, conversational |
173| Authority | Expert, authoritative | Peer-level, collaborative |
174| Emotion | Warm, empathetic | Direct, matter-of-fact |
175| Complexity | Technical, precise | Simple, accessible |
176| Energy | Bold, energetic | Calm, measured |
177| Humor | Playful, witty | Serious, earnest |
178
179Indie-hacker defaults usually land: casual, peer-level, direct, technical-but-accessible, calm, dry-humor.
180
181## Pitfalls
182
1831. **Don't be precious about the user's draft.** If something is mid, say so. Flattery hurts the writer.
184
1852. **Don't auto-rewrite without permission.** Flag issues, propose fixes, let the user decide. Their voice > your suggestion.
186
1873. **Don't over-flag on Low-severity issues.** If a post is 95% good, lead with that and surface only the High/Medium issues. Drowning the user in nitpicks is its own failure mode.
188
1894. **Calibrate severity to channel.** A "leverage" in a LinkedIn post is Medium. In a build-in-public tweet, it's High — different audiences tolerate different things.
190
1915. **Don't review your own drafts.** If `draft-post` produced the content in the same session, the writer's-block bias applies. Ask the user to read it first OR explicitly note "self-review, treat skeptically."
192
1936. **Voice drift is real.** If the user has documented voice, compare to it directly. If they've been posting for 6+ months, the voice has probably evolved — offer to re-document after 5+ reviews.
194
195## Related Skills
196
197- `draft-post` — drafting layer (the thing this skill reviews)
198- `build-in-public` — strategy layer (cadence, X Premium, infra)
199- `reply-research` — replies are a different beast; this skill is for original content