1. Understand the root cause, conditions, and exploitability of the original vulnerability.
2. Create an exact match pattern using `rg` to confirm the initial instance.
3. Identify abstraction points (function names, variable names, literal values, arguments) for generalization.
4. Iteratively generalize the pattern, changing one element at a time.
5. Review all new matches and classify them as true or false positives.
6. Stop generalizing when the false positive rate exceeds 50%.
7. Document the location, confidence, exploitability, and priority of each match.