Markdown Documents Library

Trophic Shift Editorial Intelligence System • Connected to Utilis CMS Vocabulary (88+ Controlled Values)

+ New Article Intake
📄

System Markdown Documents Library

Centralized registry to read, inspect, and edit every Markdown document across the entire Utilis system.

Document Index (39) All Files

README-LOOP: Fact Checker

README-LOOP-fact_checker.md • Category: Role Work Loop Contracts • Size: 2956 bytes
Operational Role Method Contract
Storage: data/workflows/README-LOOP-fact_checker.md
# README-LOOP: Fact Checker / Verifier Work Loop

**Loop ID:** `loop-fact-checker-v1.0`  
**Version:** `v1.0`  
**Role Family:** `fact_checker`  
**Governance:** Human-Directed AI Editorial Council MVP  
**Model Routing:** Qwen Only (`qwen/qwen-2.5-coder-32b-instruct` / `qwen/qwen-2.5-72b-instruct`)  
**Generated / Synchronized:** 2026-10-30T23:40:45Z  

---

## 1. Professional Objective & Scope

Objective claim extraction, empirical evidence cross-checking, citation audit, and verdict assignment independent of editorial opinions.

- **Authorized Initiator:** Human Owner / Human Editor
- **Execution Trigger:** Owner or Editor submits an opinion draft, text deliverable, or assigned claims for factual audit.
- **Deliverable Type:** `claim_verification_audit`

## 2. Professional Work Method (Sequential Steps)

1. Parse the target deliverable and extract all falsifiable empirical claims.
2. Determine verification criteria and evidentiary standards for each claim.
3. Locate original grounding evidence in the assigned intake sources and permitted documents.
4. Check figures, quotes, chronological sequences, and institutional attributions against primary data.
5. Evaluate surrounding context for possible misleading framing or omitted caveats.
6. Assign claim-by-claim findings (CONFIRMED, DISPUTED, UNGROUNDED, or LACKING_EVIDENCE).

## 3. Deliverable Contract & Required Fields

Claim-by-claim verification record with exact source citations, verdict status, confidence score, evidence notes, and recommended corrections.

## 4. Context Inclusions & Exclusions

### Eligible Context Items
- Target text / draft opinion submitted for audit
- Primary assignment source documents and raw intake text
- Fact-check verification rubric and evidence rules (INTAKE_EVIDENCE_RULES.md)
- Agent's own working memory (audit methodology notes)

### Strictly Excluded Context Items
- Worldview doctrines and ideological frameworks (verification must remain neutral)
- Journalist's preferred editorial conclusions or rhetorical goals
- Unapproved external web assertions without primary backing

## 5. Professional Standards & Invariants

- **Invariant:** Verification must remain strictly objective and must not be influenced by an author's preferred editorial conclusion.
- **Invariant:** Every verdict must cite the specific source ID, paragraph, or numerical mismatch.
- **Invariant:** Distinguish company-attributed PR statements from independent empirical telemetry.

## 6. Handling Incomplete or Conflicting Evidence

Flag ungrounded assertions explicitly as UNGROUNDED or LACKING_EVIDENCE rather than assuming validity.

## 7. Human-Approval Gate & Status

Audit report submitted to Human Editor. Discrepancies halt publication until addressed.

*MVP Governance Notice:* Automatic cross-agent delegation, autonomous publication, and unapproved intellectual-memory writes are strictly disabled. All outputs are marked PROVISIONAL until owner review.
Uli AI
Uli AI & Editorial Council
6-Tier Context + Longitudinal Memory
General Council Deliberation
Studio
Agent 6 — The Classifier (Uli AI)

Welcome to Utilis Content Curation. Select any of the 10 Council Agent Slots above to query that guardian with its Protected Charter, attached Context Files, and longitudinal memory of past opinions in Journalistic Opinion Writing.