Update to RUMBO Agent Reliability

{
  "description": "Reconcile conflicting evidence across sources while preserving provenance and uncertainty.",
  "included_files": [],
  "name": "deep-research-reconcile",
  "skill_md_contents": "---\nname: deep-research-reconcile\ndescription: Reconcile conflicting evidence across sources while preserving provenance and uncertainty.\n---\n\n# Deep Research & Reconcile\n\nUse when multiple sources disagree or current state is unclear.\n\n1. Gather primary evidence first.\n2. Rank evidence by authority and freshness.\n3. Preserve contradictions instead of overwriting them.\n4. Classify claims PROVEN, NOT_PROVEN, UNRESOLVED, HISTORICAL, CANDIDATE, or FAIL_CLOSED.\n5. Re-run only checks that can change a decision.\n6. Produce one reconciled state and identify its evidence.\n\nEvidence precedence:\n1. Fresh authoritative receipt.\n2. Exact current-head/object verification.\n3. Durable same-scope artifact.\n4. Explicit correction with provenance.\n5. Historical narrative.\n\nNever use a model assertion, UI state, old PASS, or score as proof of current execution.\n"
}