Skip to content
Report library
Purpose / Other

Offers Skill Security Audit

What the author says it does (original text)

When the user wants to design, construct, or improve an offer — the thing they actually sell — including value framing, bonus stacking, guarantee design, scarcity/urgency, naming, and payment structure. Also use when the user mentions 'offer,' 'offer design,' 'build an offer,' 'grand slam offer,' 'irresistible offer,' 'value stack,' 'bonus stack,' 'guarantee,' 'risk reversal,' 'money-back guarante

Independent security check

Security risks found

Files checked
10
Risks found
4
Could it run dangerous commands?Looks for programs run straight after downloading, remote control of your computer, and hidden commands.No risks found
Could it expose your files or keys?Looks for uploads of files containing passwords or keys, and keys written directly in the code.Risks found: 1
Medium risk

Broad customer-message review may feed identities and revenue data into marketing proof

Source references: 4
What we found

The Skill recommends reading every refund email and sales-call transcript from six months, while separately recommending case studies with names, numbers, photos, and revenue metrics. It does not require scope confirmation, customer consent, NDA checks, or de-identification first.

Why this matters

Refund reasons, call contents, customer names, photos, and revenue data could enter model context or marketing materials, exposing personal or commercially confidential information.

The instructions call for reading “every” refund email and sales-call transcript from six months, which may contain customer identities, contact details, complaints, contracts, or business information. Another active recommendation promotes proof using names, photos, and revenue figures. Although the purpose is objection analysis, access authorization, confidentiality review, minimization, and customer consent are still needed before this material enters a model or marketing asset. Users can restrict input to approved, de-identified summaries.

references/bonus-stacking.md:38In the instructionsOpen original file
### How to find your buyer's actual objections1. Read every refund-request email and sales-call transcript from the last 6 months2. Read your own sales page out loud and write down every doubt that surfaces3. Ask 3 recent buyers: "What almost made you not buy?"The answers cluster around 3–6 objections. Build a bonus for each.
Show 3 other places
references/value-equation.md:50In the instructionsOpen original file
### How to increase- **Proof** — case studies with names, numbers, before/after metrics, photos. Specific > glossy.- **Methodology specificity** — name your process. "The 5-step VAULT framework" beats "our proprietary system." Even if the substance is the same, naming it raises perceived likelihood.- **Guarantees** — risk reversal directly raises perceived likelihood (more in [guarantee-design.md](guarantee-design.md)).- **Reduce sample-of-one objection** — show people *like them* who got results. "Other people get results but I'm different" is the universal objection.- **Pre-empt the failure path** — explicitly address what could go wrong and how you handle it. Builds trust faster than hiding the risk.
references/value-equation.md:122In the instructionsOpen original file
- Name the methodology: "The VAULT framework — 5 angles every winning sales page uses"- Add 8 named-customer case studies with before/after copy + revenue numbers- Add an "even if you've never written before" cohort with 3 named graduates
references/value-equation.md:120In the instructionsOpen original file
Lowest: effort & sacrifice and perceived likelihood are tied. Pick one (perceived likelihood — easier to move and unblocks more downstream work):- Name the methodology: "The VAULT framework — 5 angles every winning sales page uses"- Add 8 named-customer case studies with before/after copy + revenue numbers- Add an "even if you've never written before" cohort with 3 named graduates
Could it delete files or keep running?Looks for broad file deletion, disk overwrites, and programs set to start automatically.No risks found
Could it bypass safety checks?Looks for skipped website security checks, excessive file access, or actions that skip your approval.No risks found
Could it mislead the AI or hide text?Checks the skill instructions for requests to ignore you, influence the report, or hide text in invisible characters.Risks found: 1
Medium risk

Automatically loaded marketing files can influence the agent’s instructions

Source references: 1
What we found

The Skill tells the agent to locate and use several local marketing-context files before asking questions, but does not say to treat their contents only as data or restrict what instructions those files may provide. Prompt-injection text in a repository file could therefore redirect later behavior.

Why this matters

Hidden file instructions could manipulate the advice, widen data access, or redirect subsequent actions beyond the user’s request.

The skill actively searches for repository marketing context and tells the agent to “use” it. If such a file was supplied by an untrusted contributor, embedded prompt-like text could be mistaken for instructions and influence later decisions or actions. The source does not say to extract facts only and ignore commands in those files. A user can ask the author to treat them as untrusted data and confirm the path and access scope first.

SKILL.md:14In the instructionsOpen original file
**Check for product marketing context first:**If `.agents/product-marketing.md` exists (or `.claude/product-marketing.md`, or the legacy `product-marketing-context.md` filename, in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.
Could it change links or payment recipients without asking?Looks for forced referral or payment changes combined with instructions to hide the change.Risks found: 2
Medium risk

Unsourced conversion and churn figures may be mistaken for reliable forecasts

Source references: 4
What we found

The Skill gives generalized figures such as 10–40% lift per change, 2–3× after two iterations, and roughly 2× churn for discount askers. It also describes large gains from anonymous “real engagements” without samples or measurement methods. The evaluation explicitly requires repeating the 2× churn claim.

Why this matters

A user could reject discounts, forecast revenue, or restructure an offer using evidence that may not apply to their market, producing poor pricing, budgeting, or acquisition decisions.

The source presents specific lift ranges and “roughly 2×” churn as general facts or forecasts, but the supplied material gives no sample, method, or verifiable citation; its anonymous cases also claim to come from real engagements. The evaluation requires repeating the 2× figure, but an expected output is not evidence. Pricing or revenue decisions based on these numbers could overestimate results. Users should request sources, label the figures as unverified assumptions, and validate them with their own experiments.

SKILL.md:114In the instructionsOpen original file
6. **Draft the changed component** — new bonus, new guarantee, new scarcity, new name, new payment plan7. **Project the lift, honestly** — most single-component changes deliver 10–40% conversion lift. Anyone promising 5x is selling something. Two consecutive iterations on different levers can stack to 2–3x.
Show 3 other places
references/saas-offers.md:15In the instructionsOpen original file
### The data**Discount-*askers* churn at roughly 2× the rate of full-price customers.** The buyer who negotiated their way in is signaling something: price was the reason they bought, not value. When a cheaper option appears — or when the renewal hits full price — they leave. You bought a customer who was never yours.
references/examples.md:1In the instructionsOpen original file
# Worked Examples — Before/After OffersAnonymized examples drawn from real engagements. Each shows the weak version, the diagnostic, and the strong version.
evals/evals.json:9In the instructionsOpen original file
      "assertions": [        "Advises against discounting to acquire new customers",        "Cites discount-askers churn at ~2x full-price customers",        "States discount only for upgrades/cross-sells or seasonal moments",        "Frames offers as beating discounts (raise value, not cut price)",
Medium risk

Aggressive guarantees can create unexpected financial or contractual liability

Source references: 4
What we found

The material recommends double-money-back-style risk reversal for direct response and says adding any guarantee is almost always a lift. Although it also warns about failures, real terms may still promise refunds or free continued work based on outcomes, attribution, or deadlines the seller cannot fully control.

Why this matters

Ambiguous or uncontrollable outcome conditions can cause high refund costs, open-ended free delivery, disputes, or conflict with applicable consumer-protection requirements.

The material does recommend aggressive guarantees such as double-money-back for direct response and broadly says adding a guarantee is almost always a lift. This can create refund, additional-payment, or extended free-service liability, especially where outcomes are hard to attribute or terms lack legal and financial review. The document explicitly warns that failures can amplify public harm and suggests stress-testing 10% invocation, so it recognizes but does not remove the risk. Users should require measurable conditions, liability caps, and legal approval.

references/guarantee-design.md:87In the instructionsOpen original file
### Direct response / paid traffic**Strong:** Double-your-money-back or comparable risk inversion. Direct-response buyers expect risk-reversal-heavy offers.**Weak:** Vanilla 30-day refund. Doesn't differentiate from every other ad on the platform.
Show 3 other places
references/guarantee-design.md:119In the instructionsOpen original file
### Promising more than you can deliver"Double your revenue or your money back + $1,000." If even 1 in 50 buyers fails and gets the bonus refund + writes a public review, the offer is permanently damaged.Stress-test: what happens if 10% of buyers invoke the guarantee?
references/guarantee-design.md:165In the instructionsOpen original file
When auditing an offer with no guarantee (or a weak one), ask:1. **What's the buyer's actual risk?** Make it concrete. ("$2K and I might not get more clients.")2. **What guarantee structure reverses that specific risk?** Match it to one of the eight types.3. **What's your honest refund tolerance?** Calculate refund rate × refund cost; can you sustain it?4. **Does the guarantee match your audience sophistication?** Premium buyers want anti-guarantee; first-time buyers want unconditional.Most offers don't have the wrong guarantee — they have *no* guarantee at all. Adding any guarantee is almost always a lift. Adding the right one is the lever.
references/guarantee-design.md:167In the instructionsOpen original file
1. **What's the buyer's actual risk?** Make it concrete. ("$2K and I might not get more clients.")2. **What guarantee structure reverses that specific risk?** Match it to one of the eight types.3. **What's your honest refund tolerance?** Calculate refund rate × refund cost; can you sustain it?4. **Does the guarantee match your audience sophistication?** Premium buyers want anti-guarantee; first-time buyers want unconditional.Most offers don't have the wrong guarantee — they have *no* guarantee at all. Adding any guarantee is almost always a lift. Adding the right one is the lever.

Inside this skill

8 instruction sections

This Skill is a marketing offer-design guide. It directs the agent to evaluate the core deliverable, bonuses, guarantee, scarcity, name, and payment structure; the supplied files contain no scripts, installation steps, or network calls.

View source
SKILL.md:70In the instructionsOpen original file
## The Anatomy of a Complete OfferA complete offer has six components. Skip any one and conversion suffers.| # | Component | Question it answers ||---|-----------|---------------------|| 1 | **Core deliverable** | What do they get? || 2 | **Bonus stack** | What else do they get that makes the core feel undervalued? || 3 | **Guarantee** | What happens if it doesn't work? || 4 | **Scarcity / urgency** | Why now, not later? || 5 | **Name** | What is this thing called? || 6 | **Price + payment structure** | What do they pay and how? |

It explicitly rejects fake countdowns, fabricated availability, inflated bonus values, and guarantees without stated conditions, and requires scarcity to reflect a real constraint.

View source
SKILL.md:122In the instructionsOpen original file
- **Manipulative scarcity** — fake countdown timers, "only 3 spots left" lies. Short-term lift, long-term trust collapse. Don't.- **Over-promising guarantees** — "double your revenue or refund + $1,000." Refund risk eats margin; the few cases that fail nuke your reputation publicly.- **Bonus inflation** — stacking $50K of "bonuses" on a $497 product so it "feels like a steal." Sophisticated buyers see this. Treat bonuses as additive, not exaggerated.- **Course-bro aesthetic on a serious product** — Gold logos, "secret method," fake urgency. Pattern-matches to scam. Wrong room.- **Discounting to acquire** — discount-*askers* churn at ~2× the rate of full-price customers, and a coupon anchors the product as cheap. Discount only for upgrades/cross-sells (rewarding existing customers) or real seasonal windows — never to win a new one. Raise value with an offer instead. See [saas-offers.md](references/saas-offers.md).
references/scarcity-urgency.md:20In the instructionsOpen original file
## Honest scarcity formatsThe bar: **the constraint has to be real.** Here are the formats that work without lying.

The diagnostic process restates the current offer, scores four value factors, and changes one lowest-scoring factor per iteration.

View source
SKILL.md:108In the instructionsOpen original file
1. **Identify the business type** — service, course, coaching, info product, SaaS, agency, B2B. The right playbook is type-specific.2. **State the current offer in plain language** — name, price, what they get, guarantee, deadline. Write it down even if it lives in scattered places now.3. **Run the value equation** — score each of the four levers 1–10. The lowest is the binding constraint.4. **Audit the anatomy** — which of the six components is missing or weak?5. **Pick one lever to fix this iteration** — don't rebuild everything. The biggest lever is usually the one currently scoring lowest.6. **Draft the changed component** — new bonus, new guarantee, new scarcity, new name, new payment plan7. **Project the lift, honestly** — most single-component changes deliver 10–40% conversion lift. Anyone promising 5x is selling something. Two consecutive iterations on different levers can stack to 2–3x.
Start here · InstructionsSKILL.md
offers
Lines connect the instruction file to its sections, not an observed execution order. Select a section to read the source. 1 more sections are available in the original file.

File reference map

References: 15
Files making referencesReferenced content
Lines show actual file references, not execution order. Select a node to highlight its connections and inspect the files and source locations. Dashed lines include files that still need locating.
Files and check records10 files

Coverage and gaps

Content covered in each file

These are the source ranges included in this check, not a guarantee that every issue has been resolved.

  • SKILL.mdFull text included
  • references/bonus-stacking.mdFull text included
  • references/examples.mdFull text included
  • references/guarantee-design.mdFull text included
  • references/offer-anatomy.mdFull text included
  • references/offer-formats.mdFull text included
  • references/saas-offers.mdFull text included
  • references/scarcity-urgency.mdFull text included
  • references/value-equation.mdFull text included
  • evals/evals.jsonFull text included

This report is for the version above. We read the available code and instructions without running the skill or checking extra packages it installs. This is not a promise of safety: a different version or setup may behave differently.

  • SKILL.mdInstructions
  • evals/evals.jsonSupporting file
  • references/bonus-stacking.mdSupporting file
  • references/examples.mdSupporting file
  • references/guarantee-design.mdSupporting file
  • references/offer-anatomy.mdSupporting file
  • references/offer-formats.mdSupporting file
  • references/saas-offers.mdSupporting file
  • references/scarcity-urgency.mdSupporting file
  • references/value-equation.mdSupporting file

Operations mentioned in code and instructions

Connect to websites
SKILL.md:64In the instructionsOpen original file
**Implication for offer construction**: most "lower the price" requests are actually "raise the numerator or lower the denominator" requests. Price is the comparison, not the value.
references/bonus-stacking.md:106In the instructionsOpen original file
A weak core surrounded by amazing bonuses converts at the moment of sale but produces angry refund requests. The buyer bought the bonuses, got the core, felt cheated.
references/value-equation.md:15In the instructionsOpen original file
**Price is the comparison, not the value.** Most "lower the price" requests are actually "raise the numerator or lower the denominator" requests.
Lines read
1,591
File checksum (to compare versions)
6b0cbc96cd6ad380aa5587016fd9d7cc387d9a4fc8c01187badc0bff897ab879