tkm:confidence

Tag agent outputs with confidence levels by assigning EXTRACTED, INFERRED, or AMBIGUOUS tags based on the presence or absence of evidence in source code, configuration, or data. This requires consistent tagging conventions across your analysis outputs.

Confidence tagging in root cause analysis labels debugging findings as EXTRACTED, INFERRED, or AMBIGUOUS with 0-1 scores. This standardized taxonomy reflects whether conclusions are directly pulled from evidence or deduced through inference.

Apply confidence scores to architecture insights by evaluating the evidence supporting each finding. Insights directly read from source code or configuration are tagged EXTRACTED, while deduced architectural patterns are tagged INFERRED.

Yes, you can use this confidence taxonomy to standardize reports across different analysis tasks. It provides a reference skill that other skills import to ensure consistent confidence tagging for code reviews, debugging, and architecture discovery.

EXTRACTED confidence levels apply to findings directly pulled from evidence in source code or configuration, while INFERRED levels apply to conclusions deduced without direct evidence. AMBIGUOUS tags findings where evidence presence is unclear.

No specific dependencies are required to implement confidence tagging on agent outputs. The skill operates independently using scripts and references to apply the EXTRACTED, INFERRED, and AMBIGUOUS taxonomy to your analysis results.