validate-findingslisted

# Validate Findings Opt-in coordinator-only LLM validation pass over a consolidated v4 report: adds AI assessment, verdict, and confidence per finding. Floats stay untouched unless the consolidator left them absent (partial producer output). NOT part of the automatic pipeline — invoke after `consolidate_reports.py finalize` (or `assemble`) when a triage-quality pass is wanted. **Argument**: `$ARGUMENTS` — path to the consolidated `report.json`. Edited in place. ## Inputs - A consolidated v4 report on disk (output of `consolidate_reports.py finalize` or `assemble`). - The producer commit (when `metadata.commit` is present) for best-effort source lookup via `git show`. ## Per-finding loop **Never pre-build an id-keyed lookup before `assemble` runs** — `assign_ids()` sorts by `overall_severity` before numbering, so which finding lands in an ID slot depends on the severity sort. Run the loop in place on the assembled report; if bulk pre-computation is unavoidable, key by a field `assemble` never mutates (`comment_id`, `thread_id`, `location`, content hash), never by `id`. For each finding without `ai_verdict`: 1. **Read context** — `description`, `recommendation`, any `code_snippets` (when absent, work from `description` alone), and optionally `git show <metadata.commit>:<path>` for the file in `location`. Skip the `git show` silently when `metadata.commit` is absent (non-git directory) or the command fails. 2. **Validate** — judge whether the finding holds against the co