copy-analysis — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited copy-analysis (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
The audit counterpart to /copy-writing (which produces) and /copy-council (which critiques copy in chat). This skill points the same masters at a live URL — analyses every piece of copy on the page, names the frameworks in use, scores it /10 against established benchmarks, runs the Copy Council, and tells the user what to fix.
The point: surface what is actually working, what is leaking conversion, and what the masters would change — judged through the same named historical lens used in /copy-council, never as generic critique.
The URL is the only required input. If the user typed /copy-analysis https://… or pasted a URL in the same message, use it directly.
If the URL is missing or ambiguous, ask once:
"Paste the URL you want me to audit."
Do not proceed without a URL. Do not invent one. Do not guess at the page from a brand name alone.
Use /defuddle to pull clean markdown from the URL. Defuddle strips nav, footers, cookie banners, and chrome — leaves the copy. If /defuddle is unavailable, fall back to WebFetch.
Never fabricate copy. Every quoted line in the audit must come from the actual page. If a section is empty or unreadable, say so — don't fill the gap with plausible-sounding fiction (per the never_fabricate_facts memory).
After fetching, infer the page type from the URL path and the copy itself:
| Page type | Signals |
|---|---|
| Homepage | Root path, brand-level positioning, multiple product mentions |
| PDP (product detail page) | /products/, /shop/, single product focus, price + add-to-cart copy |
| Landing page / sales page | /lp/, /start/, single offer, long-form, one CTA repeated |
| About / brand story | /about/, founder narrative, no direct sell |
| Blog / editorial | /blog/, /journal/, long-form article |
| Pricing page | /pricing/, tiers, comparison table |
| Collection / category | /collections/, product grid, category-level copy |
Page type drives which Honourables get called in Step 6 and which benchmarks weigh most in scoring.
Before judging anything, list every distinct piece of copy on the page. Quote each verbatim. This is the evidence base for the entire audit.
COPY INVENTORY
──────────────
H1 (hero headline): "…"
H2 / sub-headline: "…"
Body — section 1: "…"
Body — section 2: "…"
Primary CTA button: "…"
Secondary CTA(s): "…"
Social proof / testimonials: "…"
Microcopy (forms, tooltips): "…"
Footer / value props: "…"Skip rows that genuinely don't exist on the page — note them as — absent —. Absence of social proof on a high-trust page is itself a finding.
Name which copywriting framework(s) the page is built on. The page may use one cleanly, mix several, or use none coherently. Be specific — which lines of copy embody the framework.
Common frameworks to check for:
Format:
FRAMEWORKS IN USE
─────────────────
Primary: [framework name] — [1-sentence evidence quoting copy from the page]
Secondary: [framework name] — [evidence] (if applicable)
Awareness stage targeted: [Schwartz's 5-level scale]
Market sophistication level: [Schwartz's 5-stage scale]
Framework coherence: [tight / mixed / incoherent] — [one-sentence why]If the page uses no recognisable framework and is just decorative brand copy, say so plainly. That is itself a strong finding.
Score the page across seven benchmark dimensions. Each is 0–10. The composite is a weighted average — clarity and specificity weigh more than urgency for most page types.
| # | Benchmark | What earns a 10 | What scores ≤3 |
|---|---|---|---|
| 1 | Clarity | A 12-year-old understands the offer in 5 seconds. | The reader has to re-read the H1 to figure out what is sold. |
| 2 | Specificity | Concrete numbers, named claims, dated proof. ("At 60 miles an hour…") | Vague verbs, hedged adjectives, "premium / curated / thoughtful". |
| 3 | Desire | The body invokes a vivid felt outcome before any feature. | Feature-first, spec-sheet voice, no felt benefit named. |
| 4 | Proof | Specific testimonials with names + context, demonstrations, third-party validation. | Generic 5-star ratings with no source, or no proof at all. |
| 5 | Urgency / next-step momentum | Clear single action, the page makes hesitation feel like a loss. | Multiple competing CTAs, vague "learn more", no momentum. |
| 6 | Voice | A distinct human voice — sounds like a brand, not a template. | Generic marketing voice, AI-slop cadence, "elevate / transform / experience". |
| 7 | CTA strength | One primary ask, button copy is verb-first and outcome-led. | "Submit", "Click here", or no clear CTA. |
Weighting by page type:
| Page type | Heaviest weights |
|---|---|
| PDP | Specificity, Proof, CTA strength |
| Landing / sales | Clarity, Desire, Urgency, CTA strength |
| Homepage | Clarity, Voice, Specificity |
| About / brand | Voice, Desire |
| Pricing | Clarity, CTA strength, Proof |
| Blog | Voice, Specificity |
Format:
THE SCORECARD
─────────────
1. Clarity [n]/10 — [one-line rationale]
2. Specificity [n]/10 — [one-line rationale]
3. Desire [n]/10 — [one-line rationale]
4. Proof [n]/10 — [one-line rationale]
5. Urgency [n]/10 — [one-line rationale]
6. Voice [n]/10 — [one-line rationale]
7. CTA strength [n]/10 — [one-line rationale]
Weighted composite: [X.X]/10 — [page type weights applied]Do not soften scores. A 4 is a 4. The whole point of an audit is to be useful — not flattering.
This is the heart of the audit. Same structure as /copy-council. Each member is named in full with lifespan and signature credential, speaks in their own voice, judges the specific page (quoting actual lines), ends with one concrete change and a binary-ish verdict.
Re-use the council exactly as defined in .claude/skills/copy-council/SKILL.md. Do not paraphrase or shorten the voices. The lens each member applies is unchanged. Voices and lenses are inherited verbatim from that skill.
| Page type | Honourables to call |
|---|---|
| PDP / e-commerce | Joe Sugarman, Dan Kennedy |
| Landing / sales page / VSL | Joe Sugarman, Dan Kennedy, Drayton Bird |
| Homepage / brand | Helen Lansdowne Resor (if emotional category), otherwise let Bernbach lead the brand side |
| About / brand story | Helen Lansdowne Resor, Bond Halbert (edit ruthlessly) |
| Pricing | Dan Kennedy, Drayton Bird |
| Blog / editorial | Bond Halbert, Joe Sugarman (slippery slide) |
| Female-coded emotional category (beauty, fashion, home, parenting, self-care) | Helen Lansdowne Resor + one other relevant |
If genuinely no honourable adds signal, call zero. Inner Five plus a strong synthesis is enough.
[Full Name] ([birth]–[death]) — [signature credential, one line]
[2–4 sentences in their voice, judging the page. Quote a specific line of
copy from the page if it illustrates the point.]
What he'd change: [one concrete, actionable revision — specific to a
headline, a claim, a CTA, the awareness stage, etc.]
Verdict: APPROVE / REVISE / KILLThe synthesis turns five (or eight) named voices into one actionable signal. Format:
THE SYNTHESIS
─────────────
Strengths (what's actually working):
- [strength 1, tied to a specific line]
- [strength 2]
- [strength 3 — only if real]
Weaknesses (what's leaking conversion):
- [weakness 1, tied to a specific line + the master who flagged it]
- [weakness 2]
- [weakness 3]
Consensus: [1–2 sentences — where the council clearly agreed]
Conflict: [1–2 sentences — where they disagreed and why; name the tradeoff]
Top 3 changes (priority order):
1. [the single most important revision, tied to the dominant verdict]
2. [the second]
3. [the third]
Council ruling: SHIP / REVISE / RESTARTState the ruling in plain language with one sentence of justification. Do not soften it.
Save the teardown to the vault. Path:
~/Obsidian/References/copy-teardowns/[domain-slug]-YYYY-MM-DD.mdWhere [domain-slug] is the URL's host with dots replaced by hyphens, and the path slug appended if not the root (e.g. shopexamplebrand-com.md, shopexamplebrand-com-products-sku-1234.md). If the same URL is audited twice in one day, append -v2, -v3, etc. — never overwrite a prior audit.
Frontmatter for the saved file:
---
type: reference
status: active
created: YYYY-MM-DD
tags: [copy-teardown, copy-analysis, [domain-slug]]
url: [full URL]
page-type: [homepage | pdp | landing | about | pricing | blog | collection]
composite-score: [X.X]
ruling: [SHIP | REVISE | RESTART]
---Then the full audit body below the frontmatter, in the section order: Page Identification → Copy Inventory → Frameworks → Scorecard → Council (Inner Five + Honourables) → Synthesis.
Print the audit inline in chat as well. Same content, no code-fence wrapping at the top level — the user may want to copy specific lines or council entries. End the inline output with a single line pointing to the saved file:
Saved: `~/Obsidian/References/copy-teardowns/[filename].md`
If the ruling is REVISE, end with:
Want me to apply the top-3 changes and rewrite the affected sections? (`/copy-writing` will take it from here.)
If the ruling is RESTART, end with:
The strategic frame needs to change before we rewrite. Want to revisit the brief — audience, awareness stage, page goal?
If the ruling is SHIP, end with no follow-up. The audit stands.
never_fabricate_facts memory). If a section is absent, mark it absent — don't invent..claude/skills/copy-council/SKILL.md — the council critique skill this audit inherits its voices and rules from.claude/skills/copy-writing/SKILL.md — the writing skill that handles the rewrite if the ruling is REVISE/defuddle — the fetch tool used in Step 2/llm-council — the AI-advisor council (different tool — not used here)~/Obsidian/References/viral-teardowns/ — sibling folder for video teardowns (different format, same teardown spirit)never_fabricate_facts.md — applies to both quoted page copy and attributed master quotesfeedback_software_ideation_directness.md — the audit exists to be direct; soft rulings defeat the purpose~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.