karvey-investigatelisted

# Karvey Investigate — Root-Cause Analysis ## Purpose A **cross-cutting** skill of the Karvey Method: a debugging and root-cause-analysis support layer (the Debugger role, inspired by gstack `/investigate`). **It is NOT a phase of the linear pipeline**: it does not modify `spec.json:phase` nor advance the phase flow. It is invoked any time a symptom, bug or unexpected behavior shows up, and when it finishes the pipeline continues exactly where it was. **Iron Law:** NEVER apply fixes without first investigating the root cause. The investigation produces evidence and a recommendation; the fix is applied by `karvey-impl`, respecting the corresponding gates. It is **stack-agnostic**: it uses the target's real runtime (see `karvey/rules/targets.md`), whether it's Python/Azure Functions, Vue, SQL Server, Node-RED, Asterisk, etc. ## Steps 1. **Capture the exact symptom, how to reproduce it, and WHEN it started.** - Write down the observed behavior vs. the expected one, the literal error message, stack trace, exit code, and the minimal steps to reproduce. - If there is no reliable repro, get one before continuing. Without a repro there is no serious investigation. - **Date the symptom.** Get the timestamp of the report and, if possible, of the first occurrence. Ask the reporter if it is not in the ticket. This gives you the **incident window** (first occurrence → report), and steps 2 and 3 are useless without it. An investigation that never establishes a date is not an