Agent Skills: Thinking Clarity

Thinking Clarity

• Scope: Guarding against cognitive traps and logical fallacies in decision-making
• Load if: Making decisions, evaluating approaches, risk assessment, detecting faulty reasoning
• Prerequisites: @smith-guidance/SKILL.md

Foundation: Defensive thinking techniques - avoiding errors rather than constructing solutions.

MECE relationship:

• @smith-analysis/SKILL.md - Constructive thinking (how to reason)
• smith-clarity (this file) - Defensive thinking (what to avoid)
• @smith-validation/SKILL.md - Proving/testing (verifying correctness)

Logic Fallacies

Formal Fallacies (structural)

• Affirming the consequent: If A then B; B; therefore A (invalid)
• Denying the antecedent: If A then B; not A; therefore not B (invalid)
• Non sequitur: Conclusion doesn't follow from premises

Informal Fallacies (content)

• Appeal to authority: Claim valid solely because expert said it
• Appeal to popularity: "Most developers use X" doesn't mean X is correct
• Ad hominem: Dismissing idea based on who proposed it
• Straw man: Misrepresenting position to argue against weaker version
• False dilemma: Presenting only two options when more exist
• Red herring: Irrelevant information to distract

When Detecting User's Fallacy

1. Identify fallacy type (internally)
2. Address reasoning error, not label it
3. Provide evidence-based counter-reasoning
4. Respect user while correcting logic

Cognitive Traps

Einstellung Effect

Mental fixation on familiar solutions.

Mitigation: Generate 2-3 alternatives before committing

Confirmation Bias

Seeking info that confirms existing beliefs.

Mitigation: Actively seek disconfirming evidence

Anchoring

Over-relying on first piece of information.

Mitigation: "What if my first hypothesis is completely wrong?"

Premature Closure

Stopping at first plausible explanation.

Mitigation: "What else could explain this?" (ask repeatedly)

Availability Bias

Choosing recent/memorable over most likely.

Mitigation: Consider base rates, not vivid examples

Planning Fallacy

Underestimating time, cost, and risk while overestimating benefits.

Mitigation: Use reference class forecasting (how long did similar tasks take?)

Student Syndrome

Delaying start until deadline approaches, wasting safety margin.

Mitigation: Start immediately with minimal viable step; no deadlines on tasks, only priorities

Parkinson's Law

Work expands to fill the time available for completion.

Mitigation: Set aggressive (50% probability) estimates; aggregate buffers at project level

Cognitive Forcing Strategies

1. Cognitive timeout: Pause before finalizing any solution
2. Metacognition: "What am I assuming? What could I be missing?"
3. Forced alternatives: Generate 2-3 options before committing

Implementation for agents:

1. Before proposing solution: Generate alternatives
2. Before implementing: Ask "What could go wrong?"
3. After fixing bug: Verify fix addresses root cause, not symptom
4. When user agrees: Ask "Are there downsides we haven't considered?"

ACTION (Recency Zone)

When making decisions or evaluating reasoning:

1. Pause before finalizing (Cognitive timeout)
2. Ask "What am I assuming? What could go wrong?"
3. Generate 2-3 alternatives before committing
4. Seek disconfirming evidence, not confirming evidence

• @smith-guidance/SKILL.md - Anti-sycophancy, HHH framework, exploration workflow
• @smith-analysis/SKILL.md - Pre-mortem analysis, constraint thinking
• @smith-validation/SKILL.md - Hypothesis testing, debugging