fp-pattern-investigator
When to use
- •A pattern report flags an issue and you need to understand 'why' before tuning.
Inputs
- •patterns JSON file.
- •optional: output markdown path.
Workflow
- •Open the pattern JSON and pick the highest impact patterns (high count / high fp_rate).
- •For each pattern, inspect example signal_ids (raw_data, source_api, canonical_key).
- •Write a short markdown investigation: root cause hypotheses + concrete fixes.
Outputs
- •A human-readable investigation writeup (markdown).
Guardrails
- •Avoid broad suppressions unless you can bound impact (category, collector, confidence threshold).
References
- •
references/reference.md - •
docs/QUALITY_OPS_ARCHITECTURE.md