research-ledgerlisted

# Research Ledger `deep-research` runs discard their working state once a run ends — the fetched sources, the full claim list, and the per-claim verification votes exist only for the lifetime of that run. Only what a human (or a synthesis step) chooses to promote survives. This skill exists to catch what would otherwise be **thrown away**: it turns the run's own output into a durable, warehoused ledger, so a rejected claim or an unpromoted source can be revisited later instead of re-derived from scratch. This skill does **not** replace `wiki-synthesizer` — the confirmed-claims synthesis into `Knowledge/` concept pages still happens exactly as before. Think of it as a step that runs *before* that: **research run → ledger (this skill) → synthesis (wiki-synthesizer) → concept pages**. ## Hard limitation — read this before using the skill `deep-research` is a hosted skill; its internal `Workflow` script and cache are not inspectable or editable from here. This skill cannot reach into that run's internal state after the fact — it can only capture what the run's **own final output** happens to expose (source list, claims, verdicts, sometimes vote counts or reasoning). If the run's output only reports confirmed claims with no detail on what was rejected, the ledger will be equally thin — that is a ceiling on what's recoverable, not a bug in this skill. Never pad the ledger with invented vote reasoning or claim text to make it look more complete than the source material actually