| 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.
