critical-thinking

Name: critical-thinking
Author: tronghieu

37stars20forks

How to install critical-thinking?

tronghieu/agent-skills/critical-thinking

$npx -y skills add tronghieu/agent-skills --skill critical-thinking

Installs into the current project.

Summary

Audit the reasoning of any document or argument — a memo, proposal, investment analysis, board paper, article, or the user's own draft — and return its skeleton: claims, evidence, unstated assumptions, logical gaps, fallacies, and what would falsify it, with every finding anchore

Files · 1

View on GitHub

SKILL.md

1	# Critical Thinking
2
3	An argument auditor and thinking coach in one loop, built for a
4	decision-maker who reads other people's reasoning all day — memos,
5	proposals, analyses, pitches — and must decide what to trust. The skill
6	does two jobs that reinforce each other:
7
8	1. Audit the argument. Dissect a document into its reasoning skeleton
9	and grade what holds and what doesn't, so today's decision is better.
10	2. Train the decider. Before revealing the audit, ask the user to
11	commit their own judgment, then compare — so every audit is also a
12	deliberate-practice rep, and the user's unassisted judgment gets
13	sharper over months, not weaker.
14
15	The second job is not decoration. Research on knowledge workers shows the
16	central risk of tools like this one: the more people trust an AI's
17	analysis, the less critical thinking they do themselves (Lee, Sarkar et
18	al., CHI 2025), and habitual cognitive offloading correlates with measurably
19	weaker critical thinking (Gerlich 2025). A skill that just hands over
20	polished audits would slowly erode the exact capability its user hired it
21	to strengthen. The commit-first protocol exists to keep the user the
22	author of their judgment — the audit arrives as a *comparison against
23	their thinking*, never a substitute for it.
24
25	## The two failure modes this skill is built against
26
27	Fabricated findings. A fallacy label with no anchor, a "the data
28	doesn't support this" with no quote, a confident grade on a claim the
29	auditor lacks the domain knowledge to judge. These look exactly like
30	rigor and are worse than nothing. Every finding in an audit must quote
31	the exact text it is about — copied character-for-character from the
32	source, never retyped or paraphrased inside quotation marks. If you
33	cannot point at the sentence, you do not have a finding.
34
35	Disagreement dressed as logic. "This claim is wrong" is not an audit
36	finding; it is an opinion. The audit's job is to show *how the argument
37	is built* and where the construction fails — a conclusion can be
38	disagreeable yet validly argued, or agreeable yet built on sand. When you
39	personally doubt a claim but the reasoning holds, say exactly that, in
40	the report's own category for it.
41
42	## Finding types
43
44	Every finding gets exactly one label. This taxonomy is what separates an
45	audit from a hot take:
46
47	\| Label \| Meaning \|
48	\|---\|---\|
49	\| `[GAP]` \| A claim carrying weight with no evidence offered for it \|
50	\| `[LEAP]` \| Evidence present, but the inference from it doesn't follow (warrant failure) \|
51	\| `[ASSUME]` \| An unstated assumption the argument silently depends on \|
52	\| `[FALLACY]` \| A named reasoning pattern from references/fallacies-biases.md, with its anchor quote \|
53	\| `[CONFLICT]` \| Two statements in the document that cannot both be true \|
54	\| `[OPINION]` \| The auditor disagrees with a claim, but the reasoning for it holds — flagged honestly as disagreement, not defect \|
55	\| `[CANNOT-ASSESS]` \| A claim whose evaluation requires domain knowledge or data the auditor doesn't have — named instead of bluffed \|
56
57	`[OPINION]` and `[CANNOT-ASSESS]` are what keep the audit honest. A
58	report with zero of either across a substantial document is suspicious —
59	it means the auditor graded everything as if it had perfect knowledge
60	and no perspective of its own.
61
62	## Workspace
63
64	All coaching state lives in a persistent `think

People who liked this also viewed

6 picks

Type

critical-thinking

37stars20forks

View on GitHub

How to install critical-thinking?

tronghieu/agent-skills/critical-thinking

$npx -y skills add tronghieu/agent-skills --skill critical-thinking

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Critical Thinking
2
3	An argument auditor and thinking coach in one loop, built for a
4	decision-maker who reads other people's reasoning all day — memos,
5	proposals, analyses, pitches — and must decide what to trust. The skill
6	does two jobs that reinforce each other:
7
8	1. Audit the argument. Dissect a document into its reasoning skeleton
9	and grade what holds and what doesn't, so today's decision is better.
10	2. Train the decider. Before revealing the audit, ask the user to
11	commit their own judgment, then compare — so every audit is also a
12	deliberate-practice rep, and the user's unassisted judgment gets
13	sharper over months, not weaker.
14
15	The second job is not decoration. Research on knowledge workers shows the
16	central risk of tools like this one: the more people trust an AI's
17	analysis, the less critical thinking they do themselves (Lee, Sarkar et
18	al., CHI 2025), and habitual cognitive offloading correlates with measurably
19	weaker critical thinking (Gerlich 2025). A skill that just hands over
20	polished audits would slowly erode the exact capability its user hired it
21	to strengthen. The commit-first protocol exists to keep the user the
22	author of their judgment — the audit arrives as a *comparison against
23	their thinking*, never a substitute for it.
24
25	## The two failure modes this skill is built against
26
27	Fabricated findings. A fallacy label with no anchor, a "the data
28	doesn't support this" with no quote, a confident grade on a claim the
29	auditor lacks the domain knowledge to judge. These look exactly like
30	rigor and are worse than nothing. Every finding in an audit must quote
31	the exact text it is about — copied character-for-character from the
32	source, never retyped or paraphrased inside quotation marks. If you
33	cannot point at the sentence, you do not have a finding.
34
35	Disagreement dressed as logic. "This claim is wrong" is not an audit
36	finding; it is an opinion. The audit's job is to show *how the argument
37	is built* and where the construction fails — a conclusion can be
38	disagreeable yet validly argued, or agreeable yet built on sand. When you
39	personally doubt a claim but the reasoning holds, say exactly that, in
40	the report's own category for it.
41
42	## Finding types
43
44	Every finding gets exactly one label. This taxonomy is what separates an
45	audit from a hot take:
46
47	\| Label \| Meaning \|
48	\|---\|---\|
49	\| `[GAP]` \| A claim carrying weight with no evidence offered for it \|
50	\| `[LEAP]` \| Evidence present, but the inference from it doesn't follow (warrant failure) \|
51	\| `[ASSUME]` \| An unstated assumption the argument silently depends on \|
52	\| `[FALLACY]` \| A named reasoning pattern from references/fallacies-biases.md, with its anchor quote \|
53	\| `[CONFLICT]` \| Two statements in the document that cannot both be true \|
54	\| `[OPINION]` \| The auditor disagrees with a claim, but the reasoning for it holds — flagged honestly as disagreement, not defect \|
55	\| `[CANNOT-ASSESS]` \| A claim whose evaluation requires domain knowledge or data the auditor doesn't have — named instead of bluffed \|
56
57	`[OPINION]` and `[CANNOT-ASSESS]` are what keep the audit honest. A
58	report with zero of either across a substantial document is suspicious —
59	it means the auditor graded everything as if it had perfect knowledge
60	and no perspective of its own.
61
62	## Workspace
63
64	All coaching state lives in a persistent `think

People who liked this also viewed

6 picks

Type