UNPKG

@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.

28 lines 1.26 kB
/** * Factual-claim extraction for cold-path docs. * * Catches the documented failure class from `.goat-flow/learning-loop/lessons/verification.md` * ("Structural audit passing hides cold-path content drift"): README/doc prose * that claims wrong skill counts, wrong check counts, or points at files that * no longer exist. All counts compared against the code constants that are the * single source of truth. * * Scope per milestone Assumption: `README.md`, `CONTRIBUTING.md`, `docs/*.md`, * `.goat-flow/architecture.md`, `.goat-flow/code-map.md`. Lesson and footgun * files legitimately discuss historical counts in prose - excluded. */ import type { AuditContext, ContentFinding } from "./types.js"; /** * Run factual-claim checks across the configured documentation targets. * * Missing or unreadable target docs recover by skipping that file; unresolved * claims are emitted as content findings so audit can report all drift at once. * * @param ctx Audit context with target filesystem access. * @returns Factual-claim findings and number of scanned files. */ export declare function runFactualClaimChecks(ctx: AuditContext): { findings: ContentFinding[]; filesScanned: number; }; //# sourceMappingURL=check-factual-claims.d.ts.map