Qué incluye

Un SKILL.md (la guía paso a paso) + una plantilla inicial. Una vista previa:

Fraud Detector

When to use

• You're considering investing in a startup (angel round, crowdfunding, a friend's pitch)
  and want to separate evidence from storytelling before money moves.
• A company is courting you as a partner, supplier, or enterprise client
  and their claims sound impressive but unverifiable.
• You're evaluating a job offer at a "revolutionary" startup
  and need to know whether the technology behind the hype actually exists.
• A product makes extraordinary claims (medical diagnostics, energy breakthroughs, AI miracles)
  and you're deciding whether to trust it with money, health, or data.
• You read an investigation or watched a documentary
  and want a repeatable method for judging the next story that sounds too good to check.
• You're the one pitching, and you want to stress-test your own deck
  before someone else does it for you.

Workflow

1. Inventory the claims (input: pitch deck, interview, website, demo → output: claim ledger)
   Write down every factual claim the company makes, one per line:

• "Our device runs 200 tests from one drop of blood."
• "We have 40 peer-reviewed publications."
• "Revenue grew 300% last year."
• "Our AI outperforms human radiologists."
  Separate *claims about what exists* from *claims about what will exist*
  and *claims about what the technology could do in theory*.
  Do not paraphrase charitably — capture the exact words used, with the source
  (deck slide 12, keynote at 03:14, website FAQ) attached to each line.
  The whole skill depends on this ledger;
  a fuzzy inventory lets the "can vs could" trap slip straight through.
  If a claim appears in the marketing but not the technical docs, note that too —
  the gap between the two versions is often where the truth lives.

1. Dissect the semantics (input: claim ledger → output: verb audit)
   For each claim, flag the modal verbs and weasel words:
   *can, will, would, capable of, designed to, aims to, expects to, on track to*.

• "Our device can run 200 tests" is a statement about the present.
• "Our device is designed to run 200 tests" is a statement about intent.
• "Our device will run 200 tests" is a prediction, not a capability.
  Rewrite every bold claim in two columns:
  what the words literally assert, and what a listener is *supposed* to hear.
  Then check which column the evidence supports.
  If the company uses the language of the present tense to describe work
  that hasn't been done, you've found the central move of the Theranos playbook:
  present-tense verbs buying credibility for future-tense technology.
  Pay special attention to the word "can" — in fraud cases it almost always means
  "capable of, assuming the underlying science is developed,"
  which is a very different sentence from the one investors heard.

1. Test technical opacity (input: verb audit → output: opacity scorecard)
   Ask the blunt question: could a competent outsider verify any of this?
   Score each dimension 0–2 (0 = absent, 1 = partial, 2 = solid),
   with a one-line justification for every score:

• Peer review: independent, published, peer-reviewed validations —
  or only in-house whitepapers and paid-for studies?
• Mechanism detail: does the company explain *how* it works
  in enough detail to be falsifiable,
  or does it hide behind a wall of "proprietary" jargon?
• Third-party testing: has anyone unaffiliated run the thing
  and published the results, good or bad?
• Expert silence: do independent experts in the field engage with the claims,
  or is there an odd silence where professional skepticism should be?
  (Experts going quiet is itself a signal — the field often knows first.)
• Demo conditions: does the demo run live in front of you,
  or is it always a video, a guided walkthrough,
  or "not quite ready to show today"?
  A claim that scores high on ambition and zero on every verification channel
  is not a breakthrough — it's a story.
  Totals below 4 out of 10, attached to present-tense claims, are a danger zone.

1. Check financial and operational red flags (input: public filings, contracts, press → output: red-flag list)
   Run the standard sweep, and record a source for every flag:

• Founders who personally control the majority of voting shares
  with a compliant, hand-picked board and no independent directors.
• Labs or factories that journalists, partners, and investors
  are never allowed to tour.
• Ex-employees silenced by lawsuits and NDAs
  rather than claims answered on the merits.
• Revenue that comes from "strategic partnerships"
  with no disclosed terms, amounts, or deliverables.
• A founder whose personal brand (magazine covers, keynote stages, awards)
  is louder than the product's evidence base.
• Regulators and skeptics dismissed as "not getting it"
  instead of engaged with data.
• Auditors, law firms, or board members resigning
  without a convincing public explanation.
• Valuation jumps driven by announcements
  rather than by demonstrated technical milestones.
  One flag is a data point; three or more is a pattern.
  Write each flag down with its source —
  memory of "I read something about this" is not diligence.

1. Read the founder signals (input: interviews, depositions, public statements → output: behavior notes)
   Listen the way the "I've never lied" moment teaches you to:
   a founder who, under scrutiny, cannot name a single mistake, lie,
   or wrong call they've ever made is not displaying confidence —
   they're displaying a relationship with reality that no longer has an editor.
   Watch for these specific patterns, and quote the tape:

• The founder as the company's only source of truth
  ("only I understand the technology" — no one else can vouch for it).
• Contempt for regulators, journalists, and skeptics
  framed as visionary impatience.
• Loyalty tests disguised as mission
  ("you're either with the revolution or against it").
• A personal mythology that grows faster than the product
  (the dropped voice, the black turtleneck, the curated origin story).
• Promises that escalate exactly as evidence fails to arrive
  (each missed milestone is answered with a bigger claim).
  Charisma is not evidence. Inability to self-correct is anti-evidence.
  Capture exact quotes — "she seems shady" is an adjective;
  "asked to name one lie, she said she couldn't think of one" is a fact.

1. Run the diligence checklist (input: all of the above → output: verdict memo)
   Compile everything into a one-page verdict memo:

• The three strongest pieces of *independent* evidence you found.
• The three most significant *unverified* claims.
• The opacity score and the red-flag count.
• A recommendation in plain language:
  proceed, proceed with conditions, or walk away.
  Attach the claim ledger so a second reader can check your work.
  If you cannot fill the "independent evidence" section with anything real,
  the memo writes itself: no.
  Sleep on the memo before acting on it —
  excitement is the fraudster's best accomplice,
  and it works on the diligent too.

Vista previa de la plantilla

Fraud Detector Worksheet

Use this worksheet on any startup, product, or partnership before you commit money, time, or trust. Fill it in, then write the verdict.

1. Claim ledger

List every factual claim in the company's exact words. Tag each: exists (claimed as true now) / will-exist (promised) / could-exist (hypothetical).

2. Verb audit

Pick the 5 boldest claims. Rewrite each in two columns: what the words literally assert vs. what a listener is meant to hear.

Flag the modal verbs: can / will / would / capable of / designed to / aims to / expects to.

• Present-tense claims that depend on unbuilt technology: […]

3. Opacity scorecard

El paquete de la habilidad (SKILL.md + plantilla) está escrito en inglés.

Comprar el paquete completo de la habilidad — $4.99 Pro — las 511 habilidades