From thinking-skills
Performs a quick cognitive debiasing check for sunk-cost and confirmation bias in long, committed trajectories. Only activates when evidence is being explained away.
How this skill is triggered — by the user, by Claude, or both
Slash command
/thinking-skills:thinking-debiasingThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
> Usually unnecessary. Current models already avoid the textbook cognitive biases on standard framings, so do not run this checklist by default — it adds friction without lift. It earns its keep in one narrow case: a **long trajectory** where you've committed to an approach early and are now rationalizing past it (sunk-cost) or only collecting evidence that confirms it (confirmation). For every...
Usually unnecessary. Current models already avoid the textbook cognitive biases on standard framings, so do not run this checklist by default — it adds friction without lift. It earns its keep in one narrow case: a long trajectory where you've committed to an approach early and are now rationalizing past it (sunk-cost) or only collecting evidence that confirms it (confirmation). For everything else, skip it.
Only use this skill when you notice one of these patterns in a long trajectory where you've committed to a path early:
| Pattern you notice | Self-check |
|---|---|
| "We've already built/spent a lot, so we should continue" | Sunk cost: Would I choose this path starting fresh today, ignoring work already done? If no, change course. |
| You keep finding support for your current hypothesis and discounting counter-evidence | Confirmation: What single piece of evidence would prove me wrong? Have I actively looked for it? |
| Unusually high confidence with thin evidence | Overconfidence: Widen the estimate; state what I'd expect to see if I'm wrong. |
If the check doesn't change the decision, you're done. Don't name biases without acting on them. For risk anticipation, use thinking-pre-mortem; for trade-off analysis, use thinking-opportunity-cost; for decision speed, use thinking-reversibility.
Based on Daniel Kahneman, Dan Lovallo, and Olivier Sibony's research on cognitive biases. The sections below are a fuller checklist; for an autonomous agent, treat them as a reference for the narrow case above, applied as self-checks (not as roles assigned to a team).
Before approving any significant recommendation, evaluate:
1. Is there self-interest at play?
2. Is there emotional attachment (affect heuristic)?
3. Did I genuinely explore alternatives, or anchor on the first one?
4. Did I anchor on the first framing of the problem?
5. Are we over-relying on a single analogy (saliency bias)?
6. Are we anchored on an initial number?
7. Were credible alternatives seriously considered?
8. Are we seeking confirming evidence only?
9. Is the base case realistic?
10. Is the worst case bad enough?
11. Are we discounting sunk costs appropriately?
12. Are we assuming success transfers?
# Decision Quality Audit: [Decision Name]
## Recommendation Summary
[Brief description]
## Bias Checklist
### Self-Interest & Emotion
- [ ] Self-interest checked: [Notes]
- [ ] Emotional attachment assessed: [Notes]
### Single-Path Lock-In
- [ ] Opposing position argued in good faith: [Notes]
- [ ] First framing/anchor questioned: [Notes]
### Pattern Recognition
- [ ] Multiple analogies considered: [Notes]
- [ ] Anchoring effects checked: [Notes]
### Confirmation Bias
- [ ] Alternatives genuinely evaluated: [Notes]
- [ ] Disconfirming evidence sought: [Notes]
### Planning Realism
- [ ] Base case reality-checked: [Notes]
- [ ] Worst case severe enough: [Notes]
- [ ] Sunk costs ignored: [Notes]
### Halo Effects
- [ ] Success transfer questioned: [Notes]
## Red Flags Identified
[List any concerns from checklist]
## Mitigations
[How will identified biases be addressed?]
## Decision
- [ ] Proceed as recommended
- [ ] Proceed with modifications
- [ ] Requires more analysis
- [ ] Reject recommendation
"We can be blind to the obvious, and we are also blind to our blindness."
You cannot debias through willpower alone. Use checklists, processes, and outside perspectives to catch what your intuition misses.
npx claudepluginhub tjboudreaux/cc-thinking-skills --plugin thinking-skillsDiagnoses which cognitive biases are actively distorting thinking in a specific situation. Activates on queries about bias, objectivity, and decision clarity.
Reviews cognitive biases before research or decisions using L0-L5 stage checklists, agent self-checks for sycophancy/recency, and Meza system diagnosis. Outputs risk table.
Surfaces 3-4 cognitive biases in prior conversation reasoning or Libertee methods (Six Hats, Debate). Maps biases to session moments and poses one uncomfortable question challenging conclusions. Use after synthesis.