using-the-ledgerlisted

# Using the Ledger The decisions ledger is the single source of truth the whole package runs on. This is the short "how to use it" for any task; the authority is `references/core/ledger.md` (schema `references/core/decisions-ledger-spec.md`). ## The rules that matter - One `ledger.json`; the map, interview, and brainstorm hold no state — they project a view over it. - A `Pin` is a delta `gap = diff(to_be, as_is)`, discriminated by `kind`. `to_be` is DERIVED from the user's elected decision — never authored from code. - **Only the interview commits.** No agent sets `state: decided` or writes a `DecisionEvent`; the brainstorm writes `proposals[]` only. The human elects. - Every `DecisionEvent` carries a `flip_criteria` (when to reopen) — essential for decisions made on thin information. - Compress questions with the funnel (`references/core/interview-funnel.md`): cluster → policy → exception → proposed-default; `blocker`/`high` never go to silent default. ## Use it to - read the current pins/decisions before acting; - add a finding as a pin (with `confidence`/`provenance`; deterministic static findings carry `extracted` and skip fp-check); - record an elected decision (interview only) with a `flip_criteria`; - feed the feedback loop (`references/core/feedback-loop.md`): a fired `flip_signal` reopens a pin. ## Operate it through the runtime — never by hand Every rule above is enforced in code. A hand-written pin in `ledger.json` bypasses all of them **silently**: