Glossary

Glossary

This page provides technical and tradecraft definitions for the Claude Sleuth ecosystem. It covers codebase entities, intelligence methodology, and the data standards enforced across the 56-task pipeline.

System Architecture & Code Entities

The following diagram maps high-level system concepts to their specific implementation in the codebase.

Natural Language to Code Entity Mapping

graph TD
    subgraph "Natural Language Space"
        A["Investigation Task"]
        B["Case Persistence"]
        C["Analytical Frameworks"]
        D["Intelligence Cycle"]
    end

    subgraph "Code Entity Space"
        A1["scripts/task_runner.py"]
        A2[".sleuth-progress.json"]
        B1["server/schema.sql"]
        B2["D1 Database (SQLite)"]
        C1["scripts/source_grader.py"]
        C2["scripts/entity_resolver.py"]
        D1["scripts/config.py"]
    end

    A -- "orchestrated by" --> A1
    A1 -- "tracks state in" --> A2
    B -- "defined in" --> B1
    B1 -- "instantiated as" --> B2
    C -- "implements 6x6" --> C1
    C -- "implements Fellegi-Sunter" --> C2
    D -- "metadata defined in" --> D1

Loading

---

Core Technical Terms

Term	Definition	Implementation / File Reference
ACH	Analysis of Competing Hypotheses. A methodology for evaluating multiple explanations for a set of data by seeking to disprove them rather than confirm them.
Admiralty 6x6	A matrix system for grading source reliability (A-F) and information credibility (1-6) independently.	,
CSDb	Claude Sleuth Database. The persistent storage layer backed by Cloudflare D1.	,
CSP	Cognitive Surrogate Profile. A 16-section psychological profile of a subject built via the PSS MCP.
Fellegi-Sunter	A mathematical framework for probabilistic record linkage used to resolve entities across different data sources.
ICD 203	Intelligence Community Directive 203. Standards for analytic rigor, specifically governing probabilistic language (e.g., "Almost Certain" > 95%).	,
MCP	Model Context Protocol. The interface used by Claude to interact with external tools like CSDb, Thinking Toolkit, and PSS.
PLAN	Proportionality, Legality, Accountability, and Necessity. The justification framework required before any collection vector is executed.	,
POLE	Person, Object, Location, Event. The core data schema used for structuring extracted intelligence.	,
STEEPLES	Social, Technical, Economic, Environmental, Political, Legal, Ethical, Safety. A framework for environmental assessment during Phase 1.

---

Data Flow & Pipeline Mechanics

The investigation moves through 6 phases, managed by task_runner.py. Each task requires specific scripts and populates specific tables in the CSDb.

Task Execution and Data Persistence Flow

sequenceDiagram
    participant Analyst as "Claude / Analyst"
    participant TR as "scripts/task_runner.py"
    participant Script as "scripts/[domain]_intel.py"
    participant CSDb as "Cloudflare D1 (CSDb)"

    Analyst->>TR: next
    TR->>Analyst: Task Card (Scripts, Templates, MCPs)
    Analyst->>Script: python3 scripts/xxx.py --args
    Script-->>Analyst: Structured JSON/Output
    Analyst->>CSDb: add_entity / register_evidence
    Analyst->>TR: done
    TR->>CSDb: save_progress
    TR-->>Analyst: Transition to Next Task

Loading

---

Intelligence Tradecraft Jargon

Admiralty 6x6 Matrix

The system used in source_grader.py to prevent "workspace poisoning."

• Reliability (A-F): Grades the source's history (e.g., A = Completely reliable, F = Reliability cannot be judged).
• Credibility (1-6): Grades the specific claim (e.g., 1 = Confirmed by other sources, 6 = Truth cannot be judged).
• Source:, README.md:66.

Analysis of Competing Hypotheses (ACH)

Implemented in Phase 5 (hypcog).

• Inconsistency Principle: The core logic where the analyst seeks to reject hypotheses. The surviving hypothesis is the one with the least evidence against it, not the most for it.
• Diagnosticity: The degree to which a piece of evidence helps distinguish one hypothesis from another.
• Source:, README.md:67.

ICD 203 Probabilistic Scale

Standardized language used in report_generator.py.

• Almost Certain: >95%
• Very Likely: 80-95%
• Likely: 55-80%
• Roughly Even: 45-55%
• Unlikely: 20-45%
• Very Unlikely: 5-20%
• Almost No Chance: <5%
• Source:.

POLE Schema

The structural foundation for the entities and relationships tables in server/schema.sql.

• Person: Identity records, aliases, and biographics.
• Object: Physical or digital assets (e.g., vehicles, domains, accounts).
• Location: Geospatial coordinates and addresses.
• Event: Occurrences tied to a specific UTC timestamp.
• Source:,.

---

Script Categories

Scripts are modular and grouped by the investigative domain they serve.

Category	Scripts
Core & Grading	task_runner.py, template_builder.py, source_grader.py
Identity & Corporate	entity_resolver.py, corporate_intel.py, username_enum.py, sanctions_screen.py
Network & Technical	domain_intel.py, network_graph.py
Evidence & Reporting	evidence_preservation.py, content_archiver.py, report_generator.py
Temporal & Geo	chronological_matrix.py, geolocation.py