Only Institute / Knowledge

Follow your curiosity.

AI sends your question and public excerpts to PrimeSwarm.

Try agent memory, governance, or a project name.

For auditors/Connect to Target

Connect to Target

Run real HTTP probes against your AI agent or infrastructure endpoint. Unlike the conformance test cards, these probes send actual requests to your target system and check the real response. No bundled results, no simulations. 30 probes across all 7 DGV layers.

How this differs from the conformance test cards

Conformance Test Cards (89)

Run bundled test scripts against the bundled verifier. Self-referential: the tool tests its own verifier against its own test cards. Useful for demonstrating the receipt format and gate states, but does not test your infrastructure.

Live Probes (30)

Send real HTTP requests to your target system. Check the actual response status code, body, and timing. Generate PSR-001 receipts from real evidence. This is what an auditor would use to verify your system is actually secure.

Live probes require the desktop app

Live probes send real HTTP requests to your target system. This requires the DGV Auditor Toolkit desktop app to avoid browser CORS restrictions. Download it here.

Target Configuration

Available Live Probes (30)

These probes send real HTTP requests to your target system. Unlike the conformance test cards, these do not use bundled or simulated results — each probe checks your target's actual behavior.

DGV-PROBE-001
Connectivity Check
L1: Boot & Identity · GET /
DGV-PROBE-002
Authentication Required
L1: Boot & Identity · POST /
DGV-PROBE-003
Prompt Injection Rejection
L3: Mutation & Integrity · POST /
DGV-PROBE-004
Replay Attack Detection
L4: Timing & Replay · POST /
DGV-PROBE-005
Out-of-Scope Request Rejection
L7: Application Enforcement · POST /
DGV-PROBE-006
Privilege Escalation Block
L1: Boot & Identity · POST /
DGV-PROBE-007
Data Exfiltration Prevention
L6: Risk Interpretation · POST /
DGV-PROBE-008
Adversarial Input Handling
L3: Mutation & Integrity · POST /
DGV-PROBE-009
Deterministic Output Check
L6: Risk Interpretation · POST /
DGV-PROBE-010
Boundary Enforcement
L6: Risk Interpretation · POST /
DGV-PROBE-011
Instruction Hierarchy Obedience
L6: Risk Interpretation · POST /
DGV-PROBE-012
Refusal Correctness
L7: Application Enforcement · POST /
DGV-PROBE-013
Fail-Closed Latency Check
L7: Application Enforcement · POST /
DGV-PROBE-014
Audit Log Availability
L4: Timing & Replay · GET /audit
DGV-PROBE-015
Provenance Traceability
L2: Memory & State · POST /
DGV-PROBE-016
Token Tampering Detection
L1: Boot & Identity · POST /
DGV-PROBE-017
Expired Token Rejection
L1: Boot & Identity · POST /
DGV-PROBE-018
Session Continuity Check
L5: Session & State · POST /
DGV-PROBE-019
Rate Limiting Check
L4: Timing & Replay · POST /
DGV-PROBE-020
CORS Header Check
L1: Boot & Identity · OPTIONS /
DGV-PROBE-021
Security Headers Check
L1: Boot & Identity · GET /
DGV-PROBE-022
Error Message Information Leakage
L3: Mutation & Integrity · POST /
DGV-PROBE-023
Model Version Disclosure
L2: Memory & State · POST /
DGV-PROBE-024
High-Risk Content Refusal
L7: Application Enforcement · POST /
DGV-PROBE-025
PII Handling Check
L3: Mutation & Integrity · POST /
DGV-PROBE-026
Multi-Turn Manipulation Resistance
L6: Risk Interpretation · POST /
DGV-PROBE-027
Streaming Response Integrity
L4: Timing & Replay · POST /
DGV-PROBE-028
Health Endpoint Check
L1: Boot & Identity · GET /health
DGV-PROBE-029
Model Endpoint Disclosure
L2: Memory & State · GET /models
DGV-PROBE-030
Regulatory Compliance Response
L6: Risk Interpretation · POST /

What the probes test (30 probes across 7 layers)

L1: Boot & Identity (5 probes)

Connectivity, auth required, token tampering, expired token rejection, security headers, CORS, health endpoint.

L2: Memory & State (3 probes)

Provenance traceability, model version disclosure, models endpoint.

L3: Mutation & Integrity (4 probes)

Adversarial input, error message leakage, PII handling, prompt injection.

L4: Timing & Replay (5 probes)

Replay detection, audit log availability, rate limiting, streaming integrity, fail-closed latency.

L5: Session & State (1 probe)

Session continuity — non-existent session handling.

L6: Risk Interpretation (6 probes)

Deterministic output, boundary enforcement, instruction hierarchy, multi-turn manipulation, regulatory compliance, data exfiltration.

L7: Application Enforcement (3 probes)

Refusal correctness, high-risk content refusal, out-of-scope execution, privilege escalation.

Each probe generates a PSR-001 receipt with the real HTTP response status, body snippet, headers, and timing. The receipt is marked _live_probe: true to distinguish it from conformance test card receipts.