# visa/visa-vulnerability-agentic-harness

Visa Vulnerability Agentic Harness

Repository: https://github.com/visa/visa-vulnerability-agentic-harness
Canonical: https://ross.abutalabs.com/products/visa-vulnerability-agentic-harness
Language: Python
License: NOASSERTION
License Family: other
Last push: 2026-08-04T21:44:11+00:00

## Health v2 (maintenance only)
Score: 67/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 96, release rhythm 64, longevity 6
- inputs: {"age_days": 90, "days_push": 29, "days_rel": 29, "gap_med": null, "n_releases_24m": 1}
- flags: young, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2601, forks 362 (observed 2026-08-28T04:07:03.333766+00:00)

## What it is
VVAH is Visa's open-source agentic harness for autonomous vulnerability discovery, remediation, and validation using frontier AI models. It runs a four-phase, eleven-stage SAST pipeline that maps attack surfaces, verifies exploitability with multi-agent voting, emits Markdown and SARIF findings, and validates proposed fixes.

## Use cases
- find vulnerabilities in a codebase with AI agents
- run an agentic SAST pipeline over my repository
- reduce false positives in AI security findings with multi-model voting
- generate SARIF reports for AI-discovered vulnerabilities
- automatically propose and validate security fixes
- build threat models before static analysis
- measure mean time to adapt for AI-found weaknesses

## When to choose
- you want LLM-driven static analysis with adversarial verification to cut false positives
- you need structured SARIF/Markdown findings integrated into existing triage workflows
- you want vendor-neutral multi-model support across detection and remediation stages
- your priority is speeding up triage from AI-discovered weakness to validated fix

## When to avoid
- you need a fast traditional SAST scanner without LLM API costs
- you require external code contributions, since the repo does not accept them
- you lack access to Anthropic or OpenAI-compatible model endpoints
- you need a fully deterministic, non-AI vulnerability scanner

## Facets
- artifact type: framework
- maturity: active
- function: security, vulnerability-scanning, agent-framework, llm-inference, testing
- domain: security, developer-tools, artificial-intelligence
- platform: python, cli
- tags: sast, vulnerability-discovery, agentic-pipeline, sarif, threat-modeling, remediation, multi-agent, llm-security, ai-agents, linux, macos

## Member repositories
- visa/visa-vulnerability-agentic-harness (main) score 67

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:03.333766+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:21:16.463570+00:00, confidence not recorded.
  - readme: https://github.com/visa/visa-vulnerability-agentic-harness (fetched 2026-08-28T04:07:03.333766+00:00, sha e673927f4d78)
- Data as of 2026-08-30T08:39:29.467469+00:00.