@blundergoat/goat-flow
Version:
AI coding agent harness and local dashboard for Claude Code, OpenAI Codex, Google Antigravity, and GitHub Copilot - setup audits, guardrails, structured skills, deny hooks, and persistent learning loops.
21 lines • 1.12 kB
TypeScript
/**
* Per-context facts narrowing for the audit pipeline. The orchestrator extracts project facts once
* and reuses them across the aggregate and per-agent audits; this module hands each audit its own
* view so a per-agent run cannot mutate the shared batch bundle. The deep-clone choices here are the
* isolation contract - the only exception is the dashboard-summary profile, which shares stack facts
* because that profile never mutates them.
*/
import type { AgentId, ProjectFacts } from "../types.js";
import type { AuditFactProfile } from "./types.js";
/**
* Build an isolated facts view for one audit context from a batch fact bundle.
*
* @param facts - shared extracted facts reused across aggregate and per-agent audits
* @param options - optional agent/profile narrowing for the returned facts view
* @returns facts narrowed to the requested agent/profile without mutating the batch bundle
*/
export declare function createAuditFactsView(facts: ProjectFacts, options?: {
agentId?: AgentId;
factProfile?: AuditFactProfile;
}): ProjectFacts;
//# sourceMappingURL=audit-facts-view.d.ts.map