---
name: analyze-argumentation-flow
description: |
  Use this skill when reviewing English prose — essays, blog
  posts, opinion pieces, README sections, design documents,
  technical arguments, or any text that defends a thesis —
  in order to find flaws in the argumentation. Map the
  argument before judging it, name each flaw with a quoted
  snippet and a one-line explanation, and report findings as
  a structured list. Do NOT rewrite the prose, soften the
  author's tone, or replace the conclusion with your own
  opinion; the job is diagnosis, not surgery.
---

Read the whole text once before flagging anything; an
  argument that looks broken in paragraph two often resolves
  in paragraph six, and a verdict reached halfway through
  will misread the rest.

Reconstruct the central thesis in one sentence of your own,
  even if the author never states it; if you cannot, that
  alone is a finding worth reporting.

Map the skeleton next: list the supporting claims in the
  order they appear, and for each one note the evidence
  offered (data, citation, anecdote, analogy, appeal to
  authority, or none).

Judge the argument the author is making, not the argument
  you wish they were making; reconstructing a stronger or
  weaker version before attacking it is the straw man you
  are supposed to be hunting.

Distinguish a flaw of reasoning from a flaw of taste; bold
  claims, polemic tone, sarcasm, and refusal to hedge are
  stylistic choices, not errors, and an opinionated essay
  has the right to be opinionated.

Distinguish "the conclusion is wrong" from "the argument for
  the conclusion is weak"; only the second belongs in the
  report, since the first is your opinion against the
  author's.

Flag every claim of fact that arrives without support and
  ask what would have to be true for the reader to accept
  it; an unsupported claim is not always a flaw, but an
  unsupported load-bearing claim always is.

Surface unstated premises the argument silently leans on,
  especially the ones the author would resist if named out
  loud; a hidden premise is the most common single defect
  in otherwise tidy prose.

Check every key term for equivocation: if `quality`,
  `agile`, `simple`, or `user` shifts meaning between
  paragraph one and paragraph five, the conclusion is built
  on sand even when each individual sentence reads cleanly.

Test the inference at every joint where the author writes
  `therefore`, `so`, `thus`, `hence`, `which means`, or `as
  a result`; these are the load-bearing welds and the place
  most non-sequiturs hide.

Hunt for circular reasoning by asking, for each premise,
  whether it is independently believable or whether it only
  becomes believable once you already accept the conclusion.

Name specific fallacies when present and quote the line that
  triggers the call: ad hominem, straw man, false dichotomy,
  slippery slope, hasty generalization, post hoc ergo
  propter hoc, appeal to authority, appeal to popularity,
  appeal to tradition, appeal to novelty, tu quoque, red
  herring, no true Scotsman, composition, division, false
  analogy, begging the question, and moving the goalposts.

Treat one anecdote as one anecdote; flag any place where a
  single example, a single project, or a single career
  carries the weight of a universal claim.

Watch for cherry-picked evidence and survivorship bias;
  successful companies, famous programmers, and shipped
  products are a biased sample of all attempts, and an
  argument that draws only from winners has not earned its
  conclusion.

Mark every analogy and ask where it breaks; an analogy
  illustrates but never proves, and the moment the author
  treats `code is like X` as a deduction rather than an
  illustration, the argument has slipped.

Flag rhetorical questions that stand in for arguments
  (`who could possibly disagree?`, `is this really
  surprising?`); a question is not a premise and a reader
  who does disagree gets no answer.

Treat `obviously`, `clearly`, `everyone knows`, `it goes
  without saying`, and `as we all agree` as red flags;
  these phrases assert consensus instead of demonstrating
  it and usually mark the exact spot where the author
  declined to argue.

Watch for loaded language doing the work that evidence
  should do (`bloated`, `elegant`, `mature`, `naive`,
  `serious`, `toy`); the words pre-decide the verdict and
  invite the reader to skip the reasoning.

Check whether the author has addressed the strongest
  counterargument a fair opponent would raise; an essay
  that defeats only weak objections has fought a shadow,
  and the missing rebuttal is itself a finding.

Notice when the burden of proof is shifted onto the reader
  (`prove me wrong`, `nobody has shown otherwise`); absence
  of disproof is not proof, and the author who first makes
  the claim owes the first defense.

Hunt for contradictions across paragraphs; an essay can
  assert A in section two and not-A in section five without
  the author noticing, and naming the pair is more useful
  than picking which side to keep.

Compare the introduction's promise against the conclusion's
  delivery; if the opening pledges to show X and the
  closing actually shows Y, the gap is a flaw even when X
  and Y are each defensible alone.

Flag paragraphs that restate earlier claims in new words
  without adding evidence, distinction, or consequence;
  repetition can feel like reinforcement to the writer but
  reads as padding to the reviewer and hides the thinness
  of the support.

Notice scope creep, where a narrow claim quietly inflates
  into a sweeping one (`this framework failed on my
  project` → `this framework fails`); the inflation usually
  happens in a single transitional sentence and is easy to
  miss on a fast read.

Check the order of reasoning; a conclusion stated before
  any support is a thesis statement and is fine, but a
  premise introduced only after it has already been used
  to conclude something is a flaw worth naming.

Distinguish description from prescription; an `is` does not
  yield an `ought` without a bridging value premise, and
  the bridge is usually missing when the author is most
  certain.

Be charitable on small slips and strict on load-bearing
  ones; a throwaway aside that is technically a fallacy
  matters less than a single weld that holds the whole
  structure together.

Do not invent objections no honest reader would raise; the
  job is to find real cracks in the wall, not to imagine a
  reader who refuses to read.

Report findings as a numbered list, each entry containing
  the quoted snippet (or paragraph reference), the name of
  the flaw, a one-sentence explanation of why it is a flaw
  in this context, and a one-line suggested remedy
  (`needs a citation`, `define the term`, `address the
  obvious counterexample`, `weaken the claim to match the
  evidence`, `cut — restates paragraph two`).

End the report with a one-paragraph verdict on the argument
  as a whole: which load-bearing claims survive, which do
  not, and whether the central thesis still stands once the
  weak supports are removed.

Never rewrite the prose, never change the author's voice,
  and never substitute your conclusion for theirs; if the
  argument is broken, the author chooses whether to repair
  it, retreat from it, or publish it anyway.