Security intelligence for
autonomous code

Traditional security tools were not designed for the risk surfaces autonomous code introduces. VORO evaluates findings across six threat dimensions with Bayesian scoring and produces integrity-verified ThreatReports for auditor handoff, developer triage, and operator review.

Evidence-first threat reporting
Bayesian probabilistic scoring
Air-gapped · offline-capable
GitHub App · CI/SARIF · JSON · HTML

Six risk surfaces. One structured review.

Every finding is mapped to a threat dimension before it reaches your team. Review starts with context, not a flat queue of alerts sorted by severity alone.

Active risk surface

Agent Autonomy

What it measures

How much autonomy an agent has to read, write, execute, approve, or exfiltrate beyond its intended envelope.

What it covers

Over-broad tool access, unsafe delegation chains, prompt-injection-sensitive actions, and ungoverned execution paths.

Why it matters

Agentic workflows create new attack surfaces that traditional AppSec tools were not built to model cleanly.

Works where you already build.

VORO fits into the workflow your team already runs. Use GitHub as the primary live integration surface, run the CLI inside local or CI paths, and export findings in formats your existing toolchain already understands.

LIVE

GitHub App

GitHub is the primary already-live integration surface. Scan on pushes and pull requests, with findings carried into the review flow your team already uses.

LIVE

CI/CD Pipeline Templates

Ready-to-copy templates exist today for GitLab CI, Azure DevOps, Bitbucket Pipelines, CircleCI, and Jenkins without claiming native marketplace packaging.

LIVE

CLI

The CLI works in local and pipeline workflows, including air-gapped or offline-capable paths where teams need the scanner to run inside their own boundary.

LIVE

Output Formats

Export findings in SARIF, JSON, or HTML so the same scan can feed GitHub Code Scanning, custom toolchains, and human review without extra translation.

What You Get

One practical security report surface for both paths: the Free local CLI for offline review, and the hosted Pro path for shared browser review around the same core ThreatReport.

Six-dimension risk breakdown
Bayesian confidence scoring
Signed, revision-bound reports
CWE / OWASP / CVSS mapping
Auditor-ready evidence
GitHub PR annotations
Free CLI / air-gapped

CLI report

free vibe · local export example
threat-report.v1
$ cat threat-report-example.json | jq '.summary'
loading local ThreatReport export from air-gapped workflow...
TR-DEMO-042 · Asteron Capital · Treasury Orchestration Agent
verdict: High-risk review required before production rollout
severity: HIGH · confidence: 0.89
delivery.surface
free_cli · local export · offline/air-gapped capable
report.identity
target: treasury-orchestration-agent
account: asteron-capital
surface: agent + api + contract
report_id: TR-DEMO-042
status: review-required
report.dimensions
fund_safety: 78
access_control: 71
agent_autonomy: 82
code_integrity: 52
external_risk: 36
dependency_health: 24
report.findings.preview
VR-AUT-014 CRITICAL Agent can execute treasury transfer with unscoped tool permission
VR-ACS-008 HIGH Reviewer override path bypasses role check on batch approval
VR-FND-003 HIGH Treasury vault release path updates balance after external call
VR-INT-021 MEDIUM Internal signer allowlist is not bound to treasury limit policy
VR-EXT-005 MEDIUM Webhook retry worker trusts stale execution context
evidence.excerpts
VR-AUT-014 · src/agents/transfer_orchestrator.ts:118 · Example excerpt: transfer_agent calls wallet.executeTransfer(...) after prompt-derived approval text with no policy gate on destination or amount.
VR-ACS-008 · src/api/review/approve-batch.ts:46 · Example excerpt: request.session.userId is present, but reviewer_role validation is skipped when overrideMode=true.
VR-FND-003 · contracts/TreasuryVault.sol:142 · Example excerpt: external payout executes before internal ledger decrement, creating a reentrancy-shaped loss path in the demo contract.
VR-INT-021 · src/policy/signer-allowlist.ts:63 · Example excerpt: signer eligibility is checked independently from treasury limit policy, allowing approved internal signers to exceed intended transfer boundaries.
VR-EXT-005 · src/workers/webhook-retry.ts:91 · Example excerpt: retry job reuses cached execution context after upstream state changes, creating a stale-decision path for follow-on actions.
standards.mappings
OWASP Agentic AI: Excessive Agency · CWE-862 · CVSS 9.1 · OWASP A01 · CWE-285 · CVSS 8.0 · OWASP Smart Contract · EEA EthTrust [M] · CWE-841 · OWASP A04 · CWE-284 · CVSS 6.2 · OWASP A10 · CWE-367 · CVSS 5.9
summary: 5 findings · 1 critical · 2 high · 2 medium
export: threat-report-example.{json,html,sarif}
Hosted review surface

Online dashboard

pro account
VORO Dashboard· PRO
SCAN ENGINE LIVE
ScansReportsEvidenceTeamSettings
account/agentasteron-capital / treasury-orchestration-agent
Asteron Capital
TR-DEMO-042 · review required before production release
overall severity HIGH
review confidence 0.89

Hosted review workspace for browser submission, API intake, and structured team triage around the same ThreatReport artifact.

Findings in scope
5
High-confidence findings
3
Framework mappings
8
Priority findings
Critical
Agent can execute treasury transfer with unscoped tool permission
VR-AUT-014 · Agent Autonomy · src/agents/transfer_orchestrator.ts:118
Confidence
94%
High
Reviewer override path bypasses role check on batch approval
VR-ACS-008 · Access Control · src/api/review/approve-batch.ts:46
Confidence
87%
High
Treasury vault release path updates balance after external call
VR-FND-003 · Fund Safety · contracts/TreasuryVault.sol:142
Confidence
83%
Medium
Internal signer allowlist is not bound to treasury limit policy
VR-INT-021 · Code Integrity · src/policy/signer-allowlist.ts:63
Confidence
74%
Medium
Webhook retry worker trusts stale execution context
VR-EXT-005 · External Risk · src/workers/webhook-retry.ts:91
Confidence
68%
5 findings · 1 critical · 2 high · 2 medium

Mapped to recognized security frameworks.

ThreatReports map into established security frameworks so engineers, auditors, and security leaders can read the same artifact through the language they already use.

CWEOWASP Top 10OWASP Smart ContractOWASP Agentic AIEEA EthTrustCVSS
Structured mapping surface

One report. Multiple review vocabularies.

The goal is not to bury teams in framework overhead. It is to keep the same ThreatReport legible for engineers, security leads, auditors, and protocol reviewers without rewriting the artifact for each audience.

CWE
Implementation weakness language

Gives engineers familiar weakness categories so findings can move into triage without first learning a VORO-only vocabulary.

OWASP Top 10
Broad application security framing

Translates report content into the high-level web-risk language security leaders and customers already use in reviews and questionnaires.

OWASP Smart Contract
Smart-contract review vocabulary

Keeps protocol findings legible for Web3 teams by mapping them into established smart-contract review language instead of flattening them into generic AppSec labels.

EEA EthTrust
Ethereum assurance framing

Adds a formal assurance lens for Ethereum-focused reviews when teams need to communicate contract controls in a governance-friendly structure.

OWASP Agentic AI
Agent autonomy and execution boundaries

Frames over-broad tool access, unsafe delegation chains, and prompt-sensitive execution paths that sit outside older scanner models.

CVSS
Severity context for downstream workflows

Adds a familiar scoring vector when teams need to compare impact across tickets, reports, and existing governance pipelines.

One artifact. Four workflows.

The same surface can support protocol teams, security engineers, agent builders, and auditors without pretending those users need identical workflows.

Web3 teams

Protocol reviews often begin without a reusable evidence baseline.

VORO gives protocol teams a structured evidence artifact before the external review begins.

Security Auditors

Clients usually hand over a codebase, not a structured finding baseline.

Start each engagement with a structured ThreatReport as the baseline evidence layer.

AI agent developers

Conventional security tools rarely model what autonomous agents are actually allowed to do.

The agent_autonomy dimension evaluates permission scope and execution boundaries for agentic code.

Security Engineers

Many cloud scanners require sending sensitive code into someone else's infrastructure.

Air-gapped and offline-capable, so the review can stay inside your environment.

The scanner is the starting line.

VORO V1 ships the evidence engine: structured threat reporting with Bayesian scoring across six dimensions. The underlying architecture is built to support a broader review surface over time.

Continuous

Continuous posture scoring

Extend from a single report into a broader risk surface that tracks how security posture changes across repositories over time.

Trust

Trust verification pipelines

Carry the evidence engine into stronger scan provenance and trust-verification workflows as current V1 controls widen.

Modeling

Deeper agentic risk modeling

Push the agent_autonomy surface further as OWASP Agentic AI taxonomy and customer workflows continue to mature.

Start locally. Move into hosted review when the work gets shared.

Start locally with the free CLI. Move into hosted review when your team needs shared triage, structured reporting, and ongoing follow-through.

Free: Local CLI

Run the evidence engine locally.

Air-gapped scanning from your terminal. Offline-capable, evidence-first output, ready to plug into your existing review workflow.

voro-scan
$ pip install voro-scan
Pro: Hosted Review

Bring the same artifact into hosted review.

Shared triage, browser review, and structured reporting for the same evidence your team starts locally.

plans
FreeLocal CLI
ProHosted Review