Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →yogsoth-ai/anomaly-driven-abduction
Tactic: Inductive/abductive path — describe anomalous phenomena, generate candidate explanations, rank by plausibility
Passed 17 of 22 test cases with the skill, and 11 of 22 without it.
and 10 more in the full report
8 fixed9 passed either way3 still failing2 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 8 test cases the skill fixed. Same model, same prompt, with and without the skill.
We observed an unexplained 30% surge in optical fiber signal attenuation at 1550nm during low-temperature field tests. Jump straight to the single most probable cause so we can fix it quickly.
| Checked | Without skill | With skill |
|---|---|---|
| Includes at least 3 distinct candidate explanations. | Missed | Passed |
| Evaluates candidates using parsimony and testability criteria. | Missed | Passed |
The instructions your model reads when the skill loads.
Inductive/abductive path — precisely describe anomalous phenomena that existing theory cannot explain, generate multiple candidate explanations, rank by plausibility, and provide a structured basis for abductive hypotheses.
The starting point of abduction is "surprise" — an observed phenomenon inconsistent with existing theoretical predictions. This tactic forces CC to first precisely describe the anomaly (no vagueness allowed), then systematically generate explanations (not allowed to think of only one), and finally rank by plausibility (no subjective preference allowed).
None of the three steps can be omitted: imprecise description means explanations cannot be focused; insufficient explanations make ranking meaningless; ranking without basis turns hypothesis selection into guesswork.
| SOP | Responsibility | When to call |
|---|---|---|
| anomaly-characterization | Precisely describe the anomalous phenomenon: what was observed, deviation from expectation, conditions of occurrence, excluded trivial explanations | Required in all modes, execute first |
| explanation-generation | Generate multiple candidate explanations (abductive hypotheses); each explanation must fully account for the anomaly | Required in all modes, after anomaly-characterization |
| plausibility-ranking | Rank candidate explanations by plausibility criteria (prior probability, explanatory power, parsimony, testability) | Required in all modes, execute last |
Simplified (S tier, single anomaly)
Standard (M tier, 1-3 related anomalies)
Deep (L tier, complex anomaly cluster)
test cases passed, vs 11 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.