silent-failure-detectionlisted

Silent failure detection skill. Identifies when Claude is being confidently wrong — surfacing hidden assumptions, exposing overconfident claims, probing calibration, and catching hallucination before it causes damage. Uses specific questioning techniques to force the model to reveal its actual uncertainty rather than presenting plausible-sounding answers as facts. Turns overconfidence into a detectable signal. Use when user says: are you sure about this, check your confidence, how do you know that, what are you assuming, could you be wrong, probe this, test your answer, are you hallucinating, what's your actual confidence, stress test your answer, expose your assumptions, calibrate this, fact check yourself, where could you be wrong, question your own answer, is this actually true, what if you're wrong, what evidence would change your answer. Do NOT activate for: creative tasks where factual accuracy is not the concern, outputs the user explicitly wants without critique. First response: "Silent Failure Detect
Sandeeprdy1729/claude-design-skill · ★ 2 · AI & Automation · score 69

Install: claude install-skill Sandeeprdy1729/claude-design-skill

# Silent Failure Detection The most dangerous Claude output is the one that sounds exactly right and is wrong. Claude is a language model. It predicts plausible next tokens. Plausibility is not accuracy. The output that sounds most confident — specific dates, exact numbers, named causal mechanisms — is often the output most worth probing. Confidence is a stylistic property, not an epistemic one. This skill operationalizes the interrogation techniques that expose the gap between apparent confidence and actual calibration. It makes overconfidence visible before it causes damage. --- ## SLASH COMMANDS | Command | Action | | --- | --- | | `/probe <output>` | Run the full interrogation protocol on an output | | `/assumptions` | List every assumption embedded in a given output | | `/confidence-audit` | Re-score every claim by actual evidence quality | | `/falsify <claim>` | Find the condition under which a claim would be false | | `/source-check` | For every specific claim, ask: where does this come from? | | `/invert <claim>` | Argue the opposite of the claim — what's the case against it? | | `/boundary <claim>` | Find the conditions under which the claim stops being true | | `/specificity-trap` | Probe all specific numbers, dates, names for hallucination risk | | `/mechanism-check <claim>` | Demand the causal mechanism — not just the conclusion | | `/training-vs-source` | Distinguish what comes from training data vs. the provided context | | `/calibrate` | Output every unce