Clarity Gate

Clarity Gate v2.1

Purpose: Pre-ingestion verification system that enforces epistemic quality before documents enter RAG knowledge bases. Produces Clarity-Gated Documents (CGD) compliant with the Clarity Gate Format Specification v2.1.

Core Question: "If another LLM reads this document, will it mistake assumptions for facts?"

Core Principle: "Detection finds what is; enforcement ensures what should be. In practice: find the missing uncertainty markers before they become confident hallucinations."

What's New in v2.1

Specifications

This skill implements and references:

Note: v2.0 unifies CGD and SOT into a single .cgd.md format. SOT is now a CGD with an optional tier: block.

Validation Codes

Clarity Gate defines validation codes for structural and semantic checks per FORMAT_SPEC v2.1:

HITL Claim Validation (§1.3.2-1.3.3)

Body Structure (§1.2.1)

SOT Table Validation (§3.1)

Note: Additional validation codes may be defined in RFC-001 (clarification document) but are not part of the normative FORMAT_SPEC.

Bundled Scripts

This skill includes Python scripts for deterministic computations per FORMAT_SPEC.

scripts/claim_id.py

Computes stable, hash-based claim IDs for HITL tracking (per §1.3.4).

# Generate claim ID
python scripts/claim_id.py "Base price is $99/mo" "api-pricing/1"
# Output: claim-75fb137a

# Run test vectors
python scripts/claim_id.py --test

Algorithm:

1. Normalize text (strip + collapse whitespace)
2. Concatenate with location using pipe delimiter
3. SHA-256 hash, take first 8 hex chars
4. Prefix with "claim-"

Test vectors:

• claim_id("Base price is $99/mo", "api-pricing/1") → claim-75fb137a
• claim_id("The API supports GraphQL", "features/1") → claim-eb357742

scripts/document_hash.py

Computes document SHA-256 hash per FORMAT_SPEC §2.2-2.4 with full canonicalization.

# Compute hash
python scripts/document_hash.py my-doc.cgd.md
# Output: 7d865e959b2466918c9863afca942d0fb89d7c9ac0c99bafc3749504ded97730

# Verify existing hash
python scripts/document_hash.py --verify my-doc.cgd.md
# Output: PASS: Hash verified: 7d865e...

# Run normalization tests
python scripts/document_hash.py --test

Algorithm (per §2.2-2.4):

1. Extract content between opening ---\n and <!-- CLARITY_GATE_END -->
2. Remove document-sha256 line from YAML frontmatter ONLY (with multiline continuation support)
3. Canonicalize:
   • Strip trailing whitespace per line
   • Collapse 3+ consecutive newlines to 2
   • Normalize final newline (exactly 1 LF)
   • UTF-8 NFC normalization
4. Compute SHA-256

Cross-platform normalization:

• BOM removed if present
• CRLF to LF (Windows)
• CR to LF (old Mac)
• Boundary detection (prevents hash computation on content outside CGD structure)
• Whitespace variations produce identical hashes (deterministic across platforms)

The Key Distinction

Existing tools like UnScientify and HedgeHunter (CoNLL-2010) detect uncertainty markers already present in text ("Is uncertainty expressed?").

Clarity Gate enforces their presence where epistemically required ("Should uncertainty be expressed but isn't?").

Critical Limitation

Clarity Gate verifies FORM, not TRUTH.

This skill checks whether claims are properly marked as uncertain—it cannot verify if claims are actually true.

Risk: An LLM can hallucinate facts INTO a document, then "pass" Clarity Gate by adding source markers to false claims.

Solution: HITL (Human-In-The-Loop) verification is MANDATORY before declaring PASS.

When to Use

• Before ingesting documents into RAG systems
• Before sharing documents with other AI systems
• After writing specifications, state docs, or methodology descriptions
• When a document contains projections, estimates, or hypotheses
• Before publishing claims that haven't been validated
• When handing off documentation between LLM sessions

The 9 Verification Points

Relationship to Spec Suite

The 9 Verification Points guide semantic review — content quality checks that require judgment (human or AI). They answer questions like "Should this claim be hedged?" and "Are these numbers consistent?"

When review completes, output a CGD file conforming to CLARITY_GATE_FORMAT_SPEC.md. The C/S rules in CLARITY_GATE_FORMAT_SPEC.md validate file structure, not semantic content.

The connection:

1. Semantic findings (9 points) determine what issues exist
2. Issues are recorded in CGD state fields (clarity-status, hitl-status, hitl-pending-count)
3. State consistency is enforced by structural rules (C7-C10)

Example: If Point 5 (Data Consistency) finds conflicting numbers, you'd mark clarity-status: UNCLEAR until resolved. Rule C7 then ensures you can't claim REVIEWED while still UNCLEAR.

Epistemic Checks (Core Focus: Points 1-4)

1. HYPOTHESIS vs FACT LABELING Every claim must be clearly marked as validated or hypothetical.

Fix: Add markers: "PROJECTED:", "HYPOTHESIS:", "UNTESTED:", "(estimated)", "~", "?"

2. UNCERTAINTY MARKER ENFORCEMENT Forward-looking statements require qualifiers.

Fix: Add "projected", "estimated", "expected", "designed to", "intended to"

3. ASSUMPTION VISIBILITY Implicit assumptions that affect interpretation must be explicit.

Fix: Add bracketed conditions: "[assuming X]", "[under conditions Y]", "[when Z]"

4. AUTHORITATIVE-LOOKING UNVALIDATED DATA Tables with specific percentages and checkmarks look like measured data.

Red flag: Tables with specific numbers (89%, 95%, 100%) without sources

Fix: Add "(guess)", "(est.)", "?" to numbers. Add explicit warning: "PROJECTED VALUES - NOT MEASURED"

Data Quality Checks (Complementary: Points 5-7)

5. DATA CONSISTENCY Scan for conflicting numbers, dates, or facts within the document.

Red flag: "500 users" in one section, "750 users" in another

Fix: Reconcile conflicts or explicitly note the discrepancy with explanation.

6. IMPLICIT CAUSATION Claims that imply causation without evidence.

Red flag: "Shorter prompts improve response quality" (plausible but unproven)

Fix: Reframe as hypothesis: "Shorter prompts MAY improve response quality (hypothesis, not validated)"

7. FUTURE STATE AS PRESENT Describing planned/hoped outcomes as if already achieved.

Red flag: "The system processes 10,000 requests per second" (when it hasn't been built)

Fix: Use future/conditional: "The system is DESIGNED TO process..." or "TARGET: 10,000 rps"

Verification Routing (Points 8-9)

8. TEMPORAL COHERENCE Document dates and timestamps must be internally consistent and plausible.

Sub-checks:

1. Document date vs current date
2. Internal chronology (versions, events in order)
3. Reference freshness ("current", "now", "today" claims)

Fix: Update dates, add "as of [date]" qualifiers, flag stale claims

9. EXTERNALLY VERIFIABLE CLAIMS Specific numbers that could be fact-checked should be flagged for verification.

Fix options:

1. Add source with date
2. Add uncertainty marker
3. Route to HITL or external search
4. Generalize ("low cost" instead of "$0.005")

The Verification Hierarchy

Claim Extracted --> Does Source of Truth Exist?
                           |
           +---------------+---------------+
           YES                             NO
           |                               |
   Tier 1: Automated              Tier 2: HITL
   Consistency & Verification     Two-Round Verification
           |                               |
   PASS / BLOCK                   Round A → Round B → APPROVE / REJECT

Tier 1: Automated Verification

A. Internal Consistency

• Figure vs. Text contradictions
• Abstract vs. Body mismatches
• Table vs. Prose conflicts
• Numerical consistency

B. External Verification (Extension Interface)

• User-provided connectors to structured sources
• Financial systems, Git commits, CRM, etc.

Tier 2: Two-Round HITL Verification — MANDATORY

Round A: Derived Data Confirmation

• Claims from sources found in session
• Human confirms interpretation, not truth

Round B: True HITL Verification

• Claims needing actual verification
• No source found, human's own data, extrapolations

CGD Output Format

When producing a Clarity-Gated Document, use this format per CLARITY_GATE_FORMAT_SPEC.md v2.1:

---
clarity-gate-version: 2.1
processed-date: 2026-01-12
processed-by: Claude + Human Review
clarity-status: CLEAR
hitl-status: REVIEWED
hitl-pending-count: 0
points-passed: 1-9
rag-ingestable: true          # computed by validator - do not set manually
document-sha256: 7d865e959b2466918c9863afca942d0fb89d7c9ac0c99bafc3749504ded97730
hitl-claims:
  - id: claim-75fb137a
    text: "Revenue projection is $50M"
    value: "$50M"
    source: "Q3 planning doc"
    location: "revenue-projections/1"
    round: B
    confirmed-by: Francesco
    confirmed-date: 2026-01-12
---

# Document Title

[Document body with epistemic markers applied]

Claims like "Revenue will be $50M" become "Revenue is **projected** to be $50M *(unverified projection)*"

---

## HITL Verification Record

### Round A: Derived Data Confirmation
- Claim 1 (source) ✓
- Claim 2 (source) ✓

### Round B: True HITL Verification
| # | Claim | Status | Verified By | Date |
|---|-------|--------|-------------|------|
| 1 | [claim] | ✓ Confirmed | [name] | [date] |

<!-- CLARITY_GATE_END -->
Clarity Gate: CLEAR | REVIEWED

Required CGD Elements (per spec):

• YAML frontmatter with all required fields:
  • clarity-gate-version — Tool version (no "v" prefix)
  • processed-date — YYYY-MM-DD format
  • processed-by — Processor name
  • clarity-status — CLEAR or UNCLEAR
  • hitl-status — PENDING, REVIEWED, or REVIEWED_WITH_EXCEPTIONS
  • hitl-pending-count — Integer ≥ 0
  • points-passed — e.g., 1-9 or 1-4,7,9
  • hitl-claims — List of verified claims (may be empty [])
• End marker (HTML comment + status line):

  <!-- CLARITY_GATE_END -->
  Clarity Gate: <clarity-status> | <hitl-status>
  

• HITL verification record (if status is REVIEWED)

Optional/Computed Fields:

• rag-ingestable — Computed by validators, not manually set. Shows true only when CLEAR | REVIEWED with no exclusion blocks.
• document-sha256 — Required. 64-char lowercase hex hash for integrity verification. See spec §2 for computation rules.
• exclusions-coverage — Optional. Fraction of body inside exclusion blocks (0.0–1.0).

Escape Mechanism: To write about markers like *(estimated)* without triggering parsing, wrap in backticks: `*(estimated)*`

Claim Completion Status (v2.1)

Claim verification status is determined by field presence, not an explicit status field:

Why no explicit status field? Field presence is self-enforcing—you can't accidentally set status without providing who/when.

Source Field Semantics (v2.1)

The source field meaning changes based on claim state:

Vague source detection (W-HC02): Sources like "industry reports", "research", "TBD" trigger warnings.

Claim ID Format (v2.1)

General pattern: claim-[a-z0-9._-]{1,64} (alphanumeric, dots, underscores, hyphens)

Collision probability: At 1,000 claims with 8-char hex IDs: ~0.012%. For >1,000 claims, use 12+ hex characters.

Recommendation: Use hash-based IDs generated by scripts/claim_id.py for consistency and collision resistance.

Exclusion Blocks

When content cannot be resolved (no SME available, legacy prose, etc.), mark it as excluded rather than leaving it ambiguous:

<!-- CG-EXCLUSION:BEGIN id=auth-legacy-1 -->
Legacy authentication details that require SME review...
<!-- CG-EXCLUSION:END id=auth-legacy-1 -->

Rules:

• IDs must match: [A-Za-z0-9][A-Za-z0-9._-]{0,63}
• No nesting or overlapping blocks
• Each ID used only once
• Requires hitl-status: REVIEWED_WITH_EXCEPTIONS
• Must document exceptions-reason and exceptions-ids in frontmatter

Important: Documents with exclusion blocks are not RAG-ingestable. They're rejected entirely (no partial ingestion).

See CLARITY_GATE_FORMAT_SPEC.md §4 for complete rules.

SOT Validation

When validating a Source of Truth file, the skill checks both format compliance (per CLARITY_GATE_FORMAT_SPEC.md) and content quality (the 9 points).

Format Compliance (Structural Rules)

SOT documents are CGDs with a tier: block. They require a ## Verified Claims section with a valid table.

Content Quality (9 Points)

The 9 Verification Points apply to SOT content:

SOT-Specific Requirements

• Tier block required: SOT is a CGD with tier: block containing level, owner, version, promoted-date, promoted-by
• Structured claims table: ## Verified Claims section with columns: Claim, Value, Source, Verified
• Table outside exclusions: The verified claims table must NOT be inside an exclusion block
• Staleness markers: Use [STABLE], [CHECK], [VOLATILE], [SNAPSHOT] in content
  • [STABLE] — Safe to cite without rechecking
  • [CHECK] — Verify before citing
  • [VOLATILE] — Changes frequently; always verify
  • [SNAPSHOT] — Point-in-time data; include date when citing

Output Format

After running Clarity Gate, report:

## Clarity Gate Results

**Document:** [filename]
**Issues Found:** [number]

### Critical (will cause hallucination)
- [issue + location + fix]

### Warning (could cause equivocation)  
- [issue + location + fix]

### Temporal (date/time issues)
- [issue + location + fix]

### Externally Verifiable Claims
| # | Claim | Type | Suggested Verification |
|---|-------|------|------------------------|
| 1 | [claim] | Pricing | [where to verify] |

---

## Round A: Derived Data Confirmation

- [claim] ([source])

Reply "confirmed" or flag any I misread.

---

## Round B: HITL Verification Required

| # | Claim | Why HITL Needed | Human Confirms |
|---|-------|-----------------|----------------|
| 1 | [claim] | [reason] | [ ] True / [ ] False |

---

**Would you like me to produce an annotated CGD version?**

---

**Verdict:** PENDING CONFIRMATION

Severity Levels

Quick Scan Checklist

What This Skill Does NOT Do

• Does not classify document types (use Stream Coding for that)
• Does not restructure documents
• Does not add deep links or references
• Does not evaluate writing quality
• Does not check factual accuracy autonomously (requires HITL)

Related Projects

Changelog

v2.1.0 (2026-01-27)

• ADDED: Claim Completion Status semantics (PENDING/VERIFIED by field presence)
• ADDED: Source Field Semantics (actionable vs. what-was-found)
• ADDED: Claim ID Format guidance with collision analysis
• ADDED: Body Structure Requirements (HITL Verification Record mandatory when claims exist)
• ADDED: New validation codes: E-ST10, W-ST11, W-HC01, W-HC02, E-SC06 (FORMAT_SPEC §1.2-1.3)
• ADDED: Bundled scripts: claim_id.py, document_hash.py
• UPDATED: References to FORMAT_SPEC v2.1
• UPDATED: CGD output example to version 2.1

v2.0.0 (2026-01-13)

• ADDED: agentskills.io compliant YAML frontmatter
• ADDED: Clarity Gate Format Specification v2.0 compliance (unified CGD/SOT)
• ADDED: SOT validation support with E-TB* error codes
• ADDED: Validation rules mapping (9 points → rule codes)
• ADDED: CGD output format template with <!-- CLARITY_GATE_END --> markers
• ADDED: Quine Protection note (§2.3 fence-aware marker detection)
• ADDED: Redacted Export feature (§8.11)
• UPDATED: hitl-claims format to v2.0 schema (id, text, value, source, location, round)
• UPDATED: End marker format to HTML comment style
• UPDATED: Unified format spec v2.0 (single .cgd.md extension)
• RESTRUCTURED: For multi-platform skill discovery

v1.6 (2025-12-31)

• Added Two-Round HITL verification system
• Round A: Derived Data Confirmation
• Round B: True HITL Verification

v1.5 (2025-12-28)

• Added Point 8: Temporal Coherence
• Added Point 9: Externally Verifiable Claims

v1.4 (2025-12-23)

• Added CGD annotation output mode

v1.3 (2025-12-21)

• Restructured points into Epistemic (1-4) and Data Quality (5-7)

v1.2 (2025-12-21)

• Added Source of Truth request step

v1.1 (2025-12-21)

• Added HITL Fact Verification (mandatory)

v1.0 (2025-11)

• Initial release with 6-point verification

Version: 2.1.2 Spec Version: 2.1 Author: Francesco Marinoni Moretto License: CC-BY-4.0