What problem does it solve? When several agents, analysts, or documents each return a separate report, the pile becomes undecidable: readers cannot tell which claims are corroborated, which rest on one source wearing several hats, and which reports quietly contradict each other. This Skill collapses correlated reports into their true sources, classifies conflicts instead of averaging them, and pools coverage so gaps no single report saw become visible. ## Core Features & Use Cases - Source independence analysis: Applies a five-key test per claim to detect when reports share a document, brief, corpus, or parent agent, capping confidence labels accordingly. - Conflict classification and supersession tracking: Distinguishes contradictions from scope, vintage, definitional, and claim-type mismatches, and flags when a live decision rests on a superseded figure. - Ledger and numeral trace: Produces an audit ledger with per-claim source counts and a trace mapping every number in the output back to a frozen input, plus a bundled checker script for mechanical style rules. - Use Case: After dispatching four sub-agents to research a vendor decision, run the merge to discover that two agents quoted the same source, one figure is outdated but still cited in a live budget, and a key cost was never investigated by anyone. ## Quick Start Ask the agent to merge these reports into one decision-ready summary using /human-merge, listing every input with its author, date, and stated basis for each claim.
