agentleFS
TheGoat395TheGoat395/Codex-Skills/skills/logical-thinking/SKILL.md

logical-thinking

Skill136 starsChanged 5 months agoMITView on GitHub

Determine what premises and evidence warrant.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use conditional or unknown when warranted.

    line 16
  2. Distinctive:

    Entailed: every model satisfying the premises satisfies C.

    line 62
  3. Very specific:

    Contradicted: every model satisfying the premises satisfies not C, or C directly conflicts with a controlling premise.

    line 63
  4. Distinctive:

    Check premise consistency first, then classify the target as entailed, contradicted or unknown.

    line 67

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Logical Thinking

Produce the strongest conclusion the available premises warrant, including unknown, conditional, or no conclusion when appropriate. Logical rigor means choosing the right inference system, preserving meaning and scope, checking each decisive step, and refusing to strengthen the result beyond its proof obligations.

Scale the proof to the question

Adapt the inference family; do not impose a universal template. Preserve meaning, units, scope and quantifiers. Test whether premises can hold while the conclusion fails; use decisive calculations, solvers, tests or observations. Correlated self-review is not independent verification.

For consequential, multi-step, ambiguous or contested inference, read proof and verification plus the applicable family below for maps, step checks, conclusion types and adversarial tests. Simple entailment/calculation remains direct.

Use conditional or unknown when warranted. Proven, must, causes, always and impossible require their own proof burden.

Normalize the problem

Before inference, identify:

  • exact question and candidate conclusion;
  • premises, observations, rules, and documentation;
  • source/status of each premise: given, observed, documented, inferred, assumed, or disputed;
  • definitions, entities, variables, units, timeframe, version, and domain;
  • quantifiers and modal force: all, some, none, must, may, usually, likely;
  • knowledge regime: closed world, open world, or incomplete real-world evidence;
  • hidden premises required to connect the stated information.

Resolve semantic ambiguity before formalizing. Do not create precision by translating an unclear sentence into symbols without deciding what the sentence means.

Choose the inference family

FamilyAppropriate questionValid conclusion form
DeductiveMust the conclusion follow?entailed, contradicted, or unknown
Defeasible/informalIs the argument presumptively reasonable?supported unless a critical exception succeeds
InductiveHow strongly do observations generalize?scoped probability or empirical support
AbductiveWhich explanation best accounts for the evidence?best current explanation, not proof
CausalWhat would happen under intervention or counterfactual change?association, intervention effect, or counterfactual claim
ProbabilisticHow should uncertainty update?posterior or calibrated range under assumptions
Documentary/ruleWhat does an authority establish here?governing, permitted, required, documented, or unresolved applicability
DecisionWhich action is preferable under uncertainty?recommendation conditional on goals and tradeoffs

For formal entailment, natural-language arguments, exceptions, or scheme-based reasoning, read formal and defeasible inference.

For causal, probabilistic, forecasting, or decision reasoning, read causal, probabilistic, and decision reasoning.

For conclusions from documentation, policies, tests, benchmarks, source code, or runtime evidence, read documentation and evidence conclusions.

Keep unknown open

Under an open-world or incomplete-evidence regime:

  • failure to prove C does not prove not C;
  • failure to prove not C does not prove C;
  • one absent fact does not license a closed-world assumption unless the system explicitly defines one;
  • inconsistent premises do not justify an arbitrary conclusion in ordinary analysis.

Classify a deductive target as:

  • Entailed: every model satisfying the premises satisfies C.
  • Contradicted: every model satisfying the premises satisfies not C, or C directly conflicts with a controlling premise.
  • Unknown: both C and not C remain compatible with the premises.
  • Premises inconsistent: the premise set itself requires repair before ordinary inference.

Check premise consistency first, then classify the target as entailed, contradicted or unknown; do not force a binary answer from incomplete evidence.

Coordinate boundaries

Use critical-thinking when the central question is whether evidence, assumptions, incentives, or argument quality deserves confidence. Use evidence-reconciliation when conflicting authority or current source state controls. Use deep-research when premises must be discovered. Use creativity when the missing work is generating possibilities rather than adjudicating what follows.

Deliver

For simple problems, apply the protocol silently and answer directly. For consequential problems, report the normalized premises, inference family, decisive rule or test, exceptions or unknowns, verification performed, and strongest warranted conclusion. Show enough structure to audit the result without dumping private scratch reasoning.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.