Score existing direct-response copy against The Donahoe Method's 8 frameworks: Four-Layer Open, Conversational Proof, Three Locks, Bullet System, Cascade Close, Greased Chute, Voice Rules + Bullshit Filter, Market Temperature. Produces a 0-100 normalized score, dimension-by-dimension findings, and prioritized rewrite recommendations.
Use when the user has existing copy (a sales page, VSL script, email, ad, opener) and wants a Method-grade audit with specific rewrite directions.
Not for net-new copy generation (use /convert sales-page or /convert open) and not for live-URL CRO teardown (use /market landing).
Instrucciones de origen · Vista previa de solo lectura
name
convert-audit
slash_command
false
pack
business-conversion
family
conversion_audit
description
Score existing direct-response copy against The Donahoe Method's 8 frameworks: Four-Layer Open, Conversational Proof, Three Locks, Bullet System, Cascade Close, Greased Chute, Voice Rules + Bullshit Filter, Market Temperature. Produces a 0-100 normalized score, dimension-by-dimension findings, and prioritized rewrite recommendations.
Use when the user has existing copy (a sales page, VSL script, email, ad, opener) and wants a Method-grade audit with specific rewrite directions.
Not for net-new copy generation (use /convert sales-page or /convert open) and not for live-URL CRO teardown (use /market landing).
triggers
["audit my copy","score my sales page","donahoe audit","method audit","audit this opener","score against the method","critique my copy","copy review against donahoe","is this copy good"]
negative_triggers
["write me new copy","generate a sales page","audit the page at this url"]
{"lineage":"The Donahoe Method (Wayland-owned operating system) - full 8-framework scoring rubric; no further lineage needed beyond the Method itself"}
Convert Audit - The Donahoe Method Audit
"Would I actually say this to someone I'm trying to help?" - The Donahoe Method, Bullshit Filter
This skill takes existing copy and scores it against all eight frameworks of The Donahoe Method. The output is a defensible scoring rubric: each dimension graded 0-10, total normalized to 0-100, plus prioritized rewrite recommendations tied to the specific framework that failed.
When to Use
Trigger phrases: "audit my copy", "score my sales page", "donahoe audit", "method audit", "audit this opener", "score against the method", "critique my copy", "copy review against donahoe", "is this copy good", /convert audit <copy-or-path>.
Use when:
The user has existing copy and wants to know what's working / what's broken
A draft has been written and needs a Method gate before shipping
A page underperforms and the user wants the diagnosis (which lock is closed? which framework failed?)
An audit is needed before a rewrite - to focus the rewrite on the right dimensions
Do NOT use for:
Generating new copy from scratch - use /convert sales-page, /convert open, etc.
Auditing a live URL's full conversion experience (page + funnel + traffic) - use /market landing <url>
Auditing the offer construction itself - use /funnel offer audit (offer ≠ copy)
Auditing visual design alone - use /convert above-fold for hero, /convert proof-stack for proof zones
The 8 Frameworks (the Method, full)
This skill scores copy on the eight Method frameworks, plus the foundation (One Person Rule) and Voice Rules + Bullshit Filter. Eight scored dimensions; total 0-80; normalized to 0-100.
The copy to audit - pasted text, file path, or asset reference. Can be a full sales page, an opener, an email, an ad, a VSL script. Audit dimensions adapt to asset type (a 200-word email won't have a Cascade Close to score; that dimension becomes N/A and the rubric reweights).
Asset type - sales page / VSL / email / ad / opener / squeeze / OTO / lead-magnet / bridge
Stated audience temperature - what the user says their audience is (Ice Cold / Cool / Warm / Hot / Boiling). Dimension 8 scores whether the copy actually matches the stated temperature.
Optional:
Stated target reader - the One Person the copy is supposed to be written to (informs Voice / Bullshit Filter scoring)
Stated offer - what's being sold (informs Want / Excuse lock scoring)
Prior Method skill outputs - if /convert open, /convert proof, etc. ran on this product, their output can be cross-checked against the copy
out_path - caller-controlled output. Defaults via build_report_path("business-conversion", instruction)
The Scoring Rubric (full, defensible)
Dimension 1 - The Open (0-10)
Framework 1: The Four-Layer Open. Layer 1 Nerve Strike + Layer 2 Side Door + Layer 3 Skin in the Game + Layer 4 Fingerprint, in 3-5 sentences.
Sub-scores (each 0-2.5; total 10):
Sub-score
What it measures
0-2.5 scale
Layer 1 - Nerve Strike
Specific, names a pain the reader already feels, no marketing-speak softeners
0=absent or vague; 1=present but generic; 2=specific; 2.5=hits the nerve cold
Layer 2 - Side Door
Casual tone, "I noticed something" feel, no pitch announcement
0=announces the pitch; 1=neutral; 2=casual; 2.5=defenses-down conversation
Layer 3 - Skin in the Game
Within first 3-4 sentences, reader knows why YOU're telling them
0=absent or credential dump; 1=hinted; 2=present with specific detail; 2.5=earned through scar/specific story
Diagnostic move: A page that "feels off" but reads competently almost always has one specific Lock closed. Identify which Lock scored lowest - that's the rewrite target.
Framework 5: Stack → Vision → Math → Safety Net → Door + P.S. + P.P.S.
Sub-scores (each 0-2):
Sub-score
What it measures
0-2 scale
Stack
Conversational value summary (not formal value-table inflation)
0=formal/inflated; 1=present; 2=conversational
Vision Close
Paints the desired state at the close
0=absent; 1=present; 2=specific & visceral
Math Close
Logical / arithmetic justification
0=absent; 1=present; 2=undeniable math
Safety Net
Strong, specific guarantee
0=absent or weak; 1=present; 2=so-strong-it's-almost-absurd
Door + P.S. / P.P.S.
Direct CTA + at least one P.S. + ideally P.P.S.
0=no clear CTA / no P.S.; 1=CTA present, no P.S.; 2=full cascade including P.S. + P.P.S.
Total = sum of 5 sub-scores (out of 10).
For asset types where the Cascade doesn't apply in full (short emails, ads, openers as standalone), score the closes that should be present and reweight. Note the reweight in the output.
This isn't Method copy. Treat as a brief and rerun the writing skills.
Workflow
Step 1 - Ingest the copy
Receive the copy via paste, file path, or reference. Confirm:
Asset type (informs which dimensions apply / N/A)
Stated audience temperature
Stated target reader (if available)
Stated offer (if available)
If asset type is "opener only" or "email" or "ad," some dimensions (Cascade Close, full Chute) get reweighted or marked N/A. Note the reweight in the output.
Step 2 - Score Dimension 1 (Open)
Read the first 3-5 sentences. Mark each sentence with its layer. Assign sub-scores. Justify each in one sentence.
Step 3 - Score Dimension 2 (Proof)
Walk every claim. Check for proof within 2-3 sentences. Catalog the proof types used. Spot-check verifiability.
Step 4 - Score Dimension 3 (Three Locks)
Identify which sentences open the Want, which build Trust, which deliver the Excuse. Score each lock independently.
Step 5 - Score Dimension 4 (Bullets)
Find every bullet block. Classify each bullet by type (Keyhole / Flip / Snapshot / Scar). Score variety, specificity, curiosity, voice.
Read aloud at speaking pace (mentally simulate). Mark stumbles, energy drops, walls of text, conclusion-endings. Score paragraph rhythm + loops + transitions + naturalness.
Grep for banned words. Check first-person + contractions. Check specificity. Run the Bullshit Filter mental check on representative passages. Apply TM-banned-phrase penalty.
Step 9 - Score Dimension 8 (Temperature)
Cross-check the copy's length, awareness assumption, tone, and pitch placement against the stated audience temperature.
Step 10 - Compute totals + prioritize fixes
Sum dimensions. Apply banned penalty. Normalize to 0-100. Identify the lowest-scoring dimensions; those drive the rewrite recommendations. Order recommendations by impact (which dimension would lift the score most if fixed?).
Step 11 - Emit the audit report
Output Template
# Donahoe Method Audit: <Assetname / topic>**Asset type:**<salespage / VSL / email / ad / opener / etc.>**Stated audience temperature:**<IceCold / Cool / Warm / Hot / Boiling>**Stated target reader:**<onespecificperson, ifprovided>**Stated offer:**<ifprovided>
---
## Headline score**Total: X / 100** - <Method-grade / Strong / Mid / Weak / Off-method / FILTER-FAILED>
| Dimension | Score | Of |
|-----------|-------|-----|
| 1. Open (Four-Layer) | X.X | 10 |
| 2. Proof (Conversational + 5 types) | X.X | 10 |
| 3. Three Locks (Want / Trust / Excuse) | X.X | 10 |
| 4. Bullets (Keyhole / Flip / Snapshot / Scar) | X.X | 10 |
| 5. Close (Cascade) | X.X | 10 |
| 6. Chute (Greased Chute) | X.X | 10 |
| 7. Voice + Bullshit Filter | X.X | 10 |
| 8. Temperature match | X.X | 10 |
| **Raw subtotal** | **X.X** | **80** |
| TM banned-phrase penalty | -X | -5 max |
| **Normalized** | **X / 100** | |
### Bullshit Filter HARD-GATE OVERRIDE (mandatory, applied AFTER normalization)
The Donahoe Method spec calls the Bullshit Filter *"the single most important question in the Method."* Without this override the rubric allows copy to score 97/100 while failing the Filter - which violates the spec.
**Rule:** if the Bullshit Filter sub-score within Dimension 7 is **0 (FAIL)**, cap the normalized score at **39 / 100** regardless of other dimension scores, and stamp the verdict **FILTER-FAILED - do NOT ship this copy until the Bullshit Filter passes.**
| Bullshit Filter sub-score | Action |
|---------------------------|--------|
| 0 (FAIL - sounds like writing, not talking) | Cap normalized score at 39/100. Tag verdict **FILTER-FAILED**. Refuse to proceed to fix-priority output until copy passes the Filter. |
| 1 (MIXED - some lines fail) | No cap. Flag as **FILTER-WARNING**. Output must include the worst 3 lines + 3 forced rewrites for each. |
| 2 (PASS - sounds like talking) | No cap. Score normally. |
Why: a beautifully-structured AI-flavored sales letter (high Open / Locks / Bullets / Close scores) but failing the Filter is *worse than a structurally weak human-written piece that passes the Filter.* The Filter is the gate to which no other dimension is allowed to be a substitute. This override aligns the rubric with the Method spec.
**The Filter sub-score must be a separate human/coach pass.** The audit skill itself cannot grade the Filter on its own copy - see `convert-bullshit-filter` for the adversarial coaching pass that produces the Filter sub-score.
---
## Dimension-by-dimension findings### Dimension 1 - Open (X / 10)- Layer 1 Nerve Strike (X / 2.5): <evidence + verdict>- Layer 2 Side Door (X / 2.5): <evidence + verdict>- Layer 3 Skin in Game (X / 2.5): <evidence + verdict>- Layer 4 Fingerprint (X / 2.5): <evidence + verdict>**What's working:**<specific>**What's broken:**<specific>**Top fix:**<action>### Dimension 2 - Proof (X / 10)- Proof rhythm (X / 2): <verdict>- 5-types coverage (X / 2): <whichtypesappear / missing>- Specificity (X / 2): <verdict + examples>- Conversational delivery (X / 2): <verdict>- Verifiability (X / 2): <flaggedanonymous / fakeitems>**What's working:**<specific>**What's broken:**<specific>**Top fix:**<action>### Dimension 3 - Three Locks (X / 10)- Lock 1 The Want (X / 3.33): <verdict>- Lock 2 The Trust (X / 3.33): <verdict>- Lock 3 The Excuse (X / 3.33): <verdict>**Closed lock:**<whichlockfailed; thisisthediagnosticpriority>**Top fix:**<action>### Dimension 4 - Bullets (X / 10)- Type variety (X / 2.5): <whichtypespresent / missing>- Specificity (X / 2.5): <evidence>- Curiosity gaps (X / 2.5): <evidence>- Voice (X / 2.5): <evidence>**What's working:**<specific>**What's broken:**<specific>**Top fix:**<action>### Dimension 5 - Close (X / 10)- Stack (X / 2): <verdict>- Vision Close (X / 2): <verdict>- Math Close (X / 2): <verdict>- Safety Net (X / 2): <verdict>- Door + P.S. + P.P.S. (X / 2): <verdict>**Missing layers:**<list>**Top fix:**<action>### Dimension 6 - Chute (X / 10)- Paragraph rhythm (X / 2.5): <verdict>- Open loops at section endings (X / 2.5): <verdict>- Transitions (X / 2.5): <verdict-mentionany "here'sthething" filler>- Read-aloud naturalness (X / 2.5): <verdict + flaggedstumbles>**Stumble points:**<listofspecificlines>**Top fix:**<action>### Dimension 7 - Voice + Bullshit Filter (X / 10)- First-person + contractions (X / 2): <verdict>- Specificity (X / 2): <verdict>- Banned-word grep (X / 2): <hitsifany>- Visceral language (X / 2): <verdict>- Bullshit Filter pass (X / 2): <verdict>**Banned-word hits:**<list, e.g., "leverage" appears 4 times>**TM banned-phrase hits (penalty):**<list>**Top fix:**<action>### Dimension 8 - Temperature match (X / 10)- Length match (X / 2.5): <verdict-copyistoolong/shortforstatedtemp?>- Awareness assumption (X / 2.5): <verdict>- Tone match (X / 2.5): <verdict>- Pitch placement (X / 2.5): <verdict>**Diagnosis:**<hot-copy-cold-audience? cold-copy-hot-audience? matched?>**Top fix:**<action>
---
## Prioritized rewrite recommendations
Ordered by impact (which fix moves the score most):
### 1. <Highest-impactfix>-**Dimension affected:**<whichone>-**Current score lift potential:**<X → Y>-**Specific action:**<concreterewritedirection>-**Sub-skill to deploy:**`/convert <skill>`### 2. <Second-highestfix>
... (3-5 total recommendations)
---
## Method audit footer
---
## Recommended next moves
- If score < 70: rerun the writing skills most affected. Examples:
- Dim 1 < 6 → `/convert open` regenerate
- Dim 3 (Lock 2) < 5 → `/convert proof` regenerate + `/convert three-locks` rebalance
- Dim 5 < 5 → `/convert close` rebuild Cascade
- Dim 6 < 5 → `/convert chute` engineer momentum + `/convert transition` rewrite section breaks
- Dim 8 < 5 → `/convert temperature` reclassify and reframe
- If score 70-84: surgical rewrites on the lowest 1-2 dimensions only; re-audit.
- If score 85+: ship it. Run `/convert audit` again post-launch on any further iteration.
Pitfalls
Scoring without context. Rate-of-conversion problems often live in the offer, not the copy. If the offer is weak, even Method-grade copy will underperform. Note the boundary: this skill audits copy, not offer construction. Refer to /funnel offer audit when offer issues are suspected.
Asset-type mismatch. A 200-word email cannot have a full Cascade Close. Don't deduct points for what doesn't apply; mark N/A and reweight transparently.
Banned-phrase false flags. Some banned phrases appear in canon references for legitimate pedagogy reasons; the banned list is for user-facing copy, not internal lineage notes. The audit deducts only for hits in the audited copy's user-facing text - quotations of the list itself do not trigger the penalty.
Subjective drift on Fingerprint. Layer 4 is the most subjective sub-score. Anchor it: would this paragraph read identical with the byline swapped to a competitor? If yes, low score.
Anchoring on a single read. Read the copy twice - once for content, once aloud at speaking pace. The second read catches Chute failures the first misses.
Confirmation bias. If the user clearly loves their copy, the temptation is to score generously. Resist. The audit's value is calibration; soft scores produce shipped weak copy.
Length-biased scoring. A short, sharp page can score 90+ if the Method elements that apply are sharp. Don't penalize brevity on asset types where brevity is correct (Boiling-temp pages, follow-up emails).
Treating the rubric as the product. The recommendations matter more than the score. Lead with the prioritized fix list; the score is the diagnostic anchor.
Lineage
This audit is built entirely on The Donahoe Method (Wayland-owned operating system). All eight dimensions trace directly to the Method's eight frameworks plus the foundation (One Person Rule), Voice Rules, and Bullshit Filter. No external lineage required beyond the Method itself.
This skill produces a defensible score. Each sub-score has explicit criteria; each dimension has a documented sub-score breakdown. Reproducibility is high if the rubric is followed.
The rubric is gradable by an external auditor - meaning a second pass of /convert audit on the same copy (different session) should produce a score within ~5 points if the rubric is followed faithfully.
The audit is upstream of /convert package's phase 5 and feeds /convert report for client-ready aggregation.
The TM banned-phrase grep is non-negotiable. Any hit in the audited copy's user-facing text triggers the penalty.
The audit can be applied iteratively: audit → rewrite → re-audit. Treat scores climbing 5-10 points per iteration as healthy progress.
The score is a diagnostic anchor, not a verdict. The prioritized fix list is the actual deliverable.
