AI agent skill · by @yogsoth-ai

Characterize Anomaly

Precisely characterize an anomaly by contrasting observation with expectation, quantifying deviation where possible, recording conditions, and excluding obvious/trivial explanations before generating hypotheses.

Archived 2026-10-23 · Source updated 2026-09-29 · License: Apache-2.0. Check the original repository for the latest version.

SKILL.md instructions

# characterize-anomaly

## Purpose

Characterize an observation that departs from an expected pattern and separate signal from measurement or protocol artifact.

## Input contract

```yaml
required: [observation, reference_pattern, condition_records]
optional: [uncertainty_estimates, replication_records, artifact_hypotheses]
constraints: [anomaly status requires a defined comparison and condition context]
```

## Procedure

1. Define the expected pattern and comparison basis.
2. Quantify the departure with uncertainty and condition alignment.
3. Test plausible data, protocol, and mechanism explanations.
4. Classify the anomaly and identify discriminating follow-up evidence.

If the anomaly is reproducible and not explained by a recording artifact, consider `generate-competing-hypotheses` as the next tactic.

## Output contract

```yaml
produces: [anomaly_description, comparison_basis, explanation_set, discriminating_evidence]
delta_fields: [findings, evidence_updates, uncertainties, recommended_jumps]
```

## Quality gates

- The reference pattern and departure measure are explicit.
- Artifact explanations are checked before causal interpretations.

## Failure and counterexamples

Do not label a rare value anomalous without a comparison distribution or ignore changed measurement conditions.

## Provenance map

- `resolved: characterize-anomaly`