run-validationlisted

Execute a validation run whose spec was signed by `reconcile`. Load `method-rules` and read the `method-rules-maintenance-rules` skill §3–§4 — the claim-transition table is the law: **`supported` is reachable ONLY through this skill's signed-window verdict.** Idea slug = $0, run_id = $1 (if missing, list signed specs in `validation-runs/` whose window is open or expired-unadjudicated, and ask). ## Preconditions (refuse with reasons if any fails) 1. `validation-runs/<run_id>-spec.md` exists, `publication_status: locked`, and its `run-signed` journal row's date is BEFORE `confirmation_window.opens_at`. A spec signed at-or-after opens_at can never produce PASS — offer to re-register a fresh window via reconcile instead. 2. No report exists yet for this run_id (reports are immutable — a re-run is a NEW run spec, never an overwrite). 3. Drift boundary: if any inbox `drift_id` exceeds `maintenance.last_reconcile.consumed_through`, STOP — reconcile first (declared drift may have invalidated this run's `applies_to`). 4. The run's `applies_to` (product/version/environment) is still identifiable against the current baseline; if drift changed the subject, the run is void — journal that and route back to reconcile. ## Execution 1. Gather evidence strictly INSIDE `confirmation_window` (`opens_at` ≤ evidence date < `closes_at`), at the rungs available (evidence grading and outward-action policy apply unchanged — outreach/deploys still need per-action approval). Evidence collected outs