/**
 * integrityFindings — the ONE rail every Context Integrity finding is filed
 * on, whichever stage detected it.
 *
 * Pattern: a tiny impure helper over tracked scope; no state of its own.
 * Role:    identity dedup across passes AND across stages. One defect emits
 *          ONE `agentfootprint.integrity.context_error` per run, however many
 *          calls, iterations or seams re-detect it.
 *
 * WHY IT IS A MODULE AND NOT A LOCAL FUNCTION. It lived inside `callLLM` while
 * every seam that filed findings lived there too, with its own header saying
 * "a second copy of this loop would eventually disagree with the first about
 * what 'already filed' means". 9.77.0 put a check at a second STAGE — the
 * write seam runs at the tool-dispatch boundary, where a result exists — and
 * a second copy is exactly what that would have meant. The rail moved instead.
 *
 * The seen-list is `scope.integrityFindingIds`, falling back to
 * `priorIntegrityFindingIds` (a resumed run's carried list) — freshest copy
 * first, since a stage may have extended the list earlier in the same pass.
 * The identity is computed with the EPOCH STRIPPED, so the same defect on a
 * later iteration is the same defect.
 */
import type { TypedScope } from 'footprintjs';
import { type ContextError } from '../../integrity/finding/types.js';
import type { AgentState } from './types.js';
/** File findings on the shared seen-list rail, one event per new identity. */
export declare function fileIntegrityFindings(scope: TypedScope<AgentState>, findings: readonly ContextError[], iteration: number): void;