Contenu du pack

Un fichier SKILL.md (le guide pratique) + un modèle de départ. Aperçu :

Deception Reader

When to use

• You're negotiating (salary, contract, partnership)
  and need to weigh whether the other side's assurances are solid or performed.
• You're interviewing a candidate, founder, or vendor
  and want to read past the polished narrative to the real track record.
• Someone's story doesn't add up — the details shift between tellings —
  and you want a structured way to test it instead of going on gut feel.
• You're conducting due diligence, a post-mortem, or a difficult conversation
  where the stakes make honesty load-bearing.
• You want to get better at reading people generally:
  meetings, pitches, press interviews, depositions.
• You're preparing for a high-stakes conversation yourself
  and want to know what your own signals will look like under pressure.

Workflow

1. Establish the baseline (input: 10–15 minutes of low-stakes conversation → output: baseline profile)
   Before you probe anything important, watch how the person talks
   when nothing is at stake:

• Their normal pace: fast, measured, or slow?
• Their gesture range: expansive, moderate, or still?
• Their eye-contact rhythm: bursts, steady, or avoidant?
• Their detail level: do small questions get rich answers or thin ones?
• Their hedging style: do they hedge often, commit hard, or mix it?
  A baseline is personal — some people always fidget,
  some always speak in absolutes, some never hold eye contact.
  You are not looking for "lying behavior";
  you are looking for *this person, acting unlike themselves*.
  Without a baseline, every observation is noise.
  The famous "I've never lied" moment only lands because the baseline expectation —
  that any adult can name at least one lie —
  makes the deviation visible.
  Write the baseline down *before* the high-stakes part begins.
  Memory rewrites the baseline to fit the conclusion; the notebook doesn't.

1. Ask for specifics, then listen to the language (input: open questions → output: language-pattern notes)
   Ask questions that require concrete detail:
   "Walk me through what happened on Tuesday."
   "Who else was in the room?"
   "What did you do first, exactly?"
   Then note the patterns — and record the quotes verbatim,
   because paraphrase loses the exact wording that carries the signal:

• Absolute denials: "I have *never* lied," "I *always* tell the truth."
  Honesty rarely needs a superlative;
  the absolute is doing the work the evidence can't.
• Inability to give a single specific example:
  deflection to generalities when a concrete instance is requested.
  "I've told so many truths I can't pick one" is the tell, not the answer.
• Over-qualifying: "To be perfectly honest," "frankly," "believe me."
  The qualifiers cluster around the sentences that need them least —
  or most.
• Distancing language: "that woman," "the situation,"
  passive constructions where "I" should be ("mistakes were made").
  Distance in grammar often tracks distance from the truth.
• Chronology gaps: the story has a detailed middle and a blurry beginning.
  Deception is usually constructed around the event, not the context —
  so the context is where the construction shows.
• Answering a different question: the response is fluent and detailed
  but subtly addresses the question you didn't ask.
  Note which pattern, which quote, and which question prompted it.

1. Watch for stress deviations (input: high-stakes questions → output: deviation log)
   When the topic turns to what matters, compare against the baseline:

• Pace changes: sudden hurry, or sudden carefulness —
  the words start racing or each one gets weighed.
• Gesture changes: hands that talked freely go still,
  or stillness breaks into fidgeting.
• Response latency: the pause before an answer that used to come instantly —
  or an answer that now comes *too* fast, rehearsed to the millisecond.
• Vocal changes: pitch rising, volume dropping, filler words appearing
  or disappearing versus their normal speech.
  Note what changed, what the baseline was, and exactly which question
  triggered it.
  One deviation is curiosity;
  a cluster of deviations around the same topic is a direction to dig.
  And deviations that *stop* the moment the topic changes
  are more informative than the ones that linger.

1. Read the non-verbals in context (input: deviation log → output: contextualized cues)
   Non-verbal cues only mean something against the baseline and the question.
   Never read them as a checklist — read them as a comparison:

• Eye behavior: breaking a normal contact rhythm,
  or sudden over-steady staring.
  Both are deviations, not opposites — compare to *their* normal,
  not to a textbook.
• Micro-expressions: a flash of an emotion that contradicts the words —
  contempt while claiming respect, fear while claiming confidence.
  They last a fraction of a second; you catch them by watching,
  not by trying.
• Self-soothing: touching face, neck, or hands
  when the topic gets specific.
  The body comforts itself when the story doesn't.
• Posture shifts: leaning away, barrier-building (arms, objects, laptop lids)
  landing exactly on the hard question.
  Weight each cue low on its own.
  The power is in convergence:
  a language pattern + a stress deviation + a non-verbal cue,
  all pointing at the same sentence.

1. Calibrate and test (input: cues → output: hypothesis + test plan)
   Turn observations into a testable hypothesis, not a verdict:
   "I think the timeline claim is the weak point,
   because the language went vague, the pace doubled,
   and the eye contact broke — all at that one question."
   Then test it — deception readers don't conclude, they *probe where the smoke is*:

• Re-ask from a different angle later in the conversation
  and compare the two versions detail by detail.
• Request a document that would corroborate the claim
  (an email, a receipt, a dated record).
• Ask an open question you already know the answer to
  and watch how the known truth is handled —
  it calibrates everything else they tell you.
  If the hypothesis survives three independent probes,
  upgrade it to a working belief.
  If a single probe clears it, drop it — holding on past that point
  is confirmation bias, not reading.

1. Write the read (input: everything → output: one-page read)
   Summarize in one page:

• The baseline, in three lines.
• The two or three strongest cue clusters —
  each with the verbatim quote, the observed deviation,
  and the question that triggered it.
• The hypothesis, in one sentence.
• What would confirm it, and what would kill it
  (the falsification condition).
• The decision the read informs —
  not "they're lying" but "don't sign until X is verified"
  or "ask Y in the follow-up."
• The mandatory caveat line:
  no single cue proves lying.
  A read that can't name its own falsification condition
  is just suspicion with formatting.
  Keep your reads. Over time they become the training data
  that makes your baseline sense — and your self-honesty — sharper.

Aperçu du modèle

Deception Reader Worksheet

Use this worksheet when someone's story doesn't add up — negotiations, interviews, diligence, hard conversations. Fill it in, then write the read.

1. Baseline profile

Spend 10–15 minutes on low-stakes topics first. Capture how they talk when nothing is at stake.

• Normal pace: [fast / moderate / slow]
• Gesture range: [expansive / moderate / still]
• Eye-contact rhythm: […]
• Detail level in answers: [rich / moderate / sparse]
• Hedging style: [hedges often / commits / mixed]
• Anything else distinctive: […]

*Without a baseline, every observation is noise.*

2. Language-pattern notes

Le pack de compétence lui-même (SKILL.md + modèle) est rédigé en anglais.

Acheter le pack complet — $4.99 Pro — les 511 compétences