Razorglint Trust Reconstruction Demos

Three local-first demos that reconstruct operational trust from existing logs, receipts, and records.

No cloud. No vendor. No dashboard. No hidden scoring.
Clone, run, verify. That's it.

---

The Problem

After something breaks — a compromise, a misconfiguration, an unauthorized change — organizations face three questions they usually cannot answer with evidence:

1. Which parts of the operational chain were still trustworthy? (not just "what was attacked")
2. Which decisions were actually authorized and evidence-based? (not just "who was logged in")
3. How long were our policies being violated before we noticed? (not just "are we compliant today")

These demos address those questions directly.

---

What's Here

Demo	Pain	Output States
Trust Continuity Mapper	Cannot prove trust held between audit points	TRUSTED · DEGRADED · UNVERIFIABLE · CONTAMINATED
Decision Provenance Tracker	Cannot prove why an agent decided, who authorized it, or what evidence was used	PROVABLE · UNPROVABLE · TAMPERED · ORPHANED
Compliance Drift Detector	Policy and system behavior diverge silently for months	ALIGNED · DRIFTING · VIOLATED · UNDECLARED

---

Run Everything

python scripts/run_all_demos.py

Expected output:

[1/3] Trust Continuity Mapper...        PASS
[2/3] Decision Provenance Tracker...    PASS
[3/3] Compliance Drift Detector...      PASS

All demos completed. Artifacts written to each demo's output/ directory.

Verify Everything

python scripts/verify_all.py

Expected output:

TrustContinuityMapper:        PASS
DecisionProvenanceTracker:    PASS
ComplianceDriftDetector:      PASS

All demo artifacts verified.

---

Requirements

• Python 3.11+
• Zero external dependencies (stdlib only)
• Any OS (Windows, macOS, Linux)
• No network required

---

How Verification Works

Every demo produces hash-linked artifacts:

1. Inputs are hashed — any modification to source data changes all downstream hashes
2. Processing chains are hash-linked — each step's output is chained to the previous
3. Outputs are sealed — the final artifact includes a hash of the complete analysis
4. Verification is independent — verify.py re-checks hashes without re-running the engine

This means: if anyone modifies any input, any intermediate state, or any output — the verification fails visibly.

---

What This Proves

• Operational trust can be reconstructed from existing event data
• Decision authorization can be traced with cryptographic linkage
• Policy drift can be measured continuously, not just at audit time
• All outputs are deterministic (same inputs → same outputs, always)
• All artifacts are independently verifiable (no trust-the-tool required)
• This works offline, locally, with zero vendor dependency

What This Does NOT Prove

• These are demos, not production systems
• They work on structured inputs — real-world log variety requires adapter work
• Classification rules are intentionally simple and visible — complex environments need tuning
• No UI, no integrations, no deployment tooling
• These narrow slices do not represent the full Razorglint governance stack
• We do not claim these are "compliant" with any framework — they produce evidence, not certifications

---

Design Principles

Principle	Meaning
Deterministic	Same inputs produce same outputs. No randomness. No model inference.
Verifiable	Every artifact can be checked independently. No "trust us."
Local-first	Runs on your machine. No cloud calls. No telemetry. No accounts.
Evidence-heavy	Claims are backed by hashable artifacts, not assertions.
No hidden scoring	Every threshold, every rule, every classification logic is in source.
Explicit boundaries	What is proven and what is not proven — stated clearly.

---

Relationship to Razorglint

These demos are intentionally narrow reproducible slices showing how the broader Razorglint proof/governance architecture approaches operational trust problems.

They are not the full stack. They are not a product you deploy. They are proof-of-method artifacts designed so you can:

1. Understand the operational problem in minutes
2. See how structured trust reconstruction works
3. Verify the outputs yourself
4. Decide if this approach fits your environment

---

Repository Structure

razorglint-trust-reconstruction-demos/
├── README.md                  ← you are here
├── QUICKSTART.md              ← 60-second setup
├── CLAIM_BOUNDARIES.md        ← what we claim, what we don't
├── MANIFEST.sha256            ← integrity check for repo contents
├── demos/
│   ├── trust-continuity-mapper/
│   ├── decision-provenance-tracker/
│   └── compliance-drift-detector/
├── scripts/
│   ├── run_all_demos.py
│   └── verify_all.py
└── sample_outputs/            ← pre-generated reference outputs

---

License

Proprietary — RazorGlint Labs. All rights reserved.

For evaluation, demonstration, and discussion purposes.
Contact for licensing, integration, or pilot conversations.