github-action

VeriSwarm Trust Check — GitHub Action

Run trust score checks, security tests, PII scans, and injection scans against your AI agents in CI.

Quick Start

name: Trust Check
on: [push, pull_request]

jobs:
  veriswarm:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: veriswarm/trust-check@v1
        with:
          api-key: ${{ secrets.VERISWARM_API_KEY }}
          agent-id: "agt_your_agent"

What It Checks

Check	Description	Fails when
Trust Score	Current agent trust score	Below min-trust-score
Decision	Trust decision for an action	Decision is deny
Security Tests	33 adversarial tests	Readiness score below fail-on-low-score
PII Scan	Scan files for personal data	PII found in scanned files
Injection Scan	Scan files for injection patterns	Injection detected

Inputs

Input	Required	Default	Description
api-key	Yes		VeriSwarm API key (use GitHub secret)
agent-id	No		Agent ID for score/test/decision checks
mode	No	all	score, test, scan, decision, or all
scan-paths	No		Glob for files to scan (e.g., **/*.py)
fail-on-deny	No	true	Fail if trust decision is deny
fail-on-injection	No	true	Fail if injection detected
fail-on-low-score	No	0	Min security readiness score (0=disabled)
min-trust-score	No	0	Min composite trust score (0=disabled)

Examples

Score check only

- uses: veriswarm/trust-check@v1
  with:
    api-key: ${{ secrets.VERISWARM_API_KEY }}
    agent-id: "agt_my_agent"
    mode: score
    min-trust-score: 60

PII scan on Python files

- uses: veriswarm/trust-check@v1
  with:
    api-key: ${{ secrets.VERISWARM_API_KEY }}
    mode: scan
    scan-paths: "**/*.py,**/*.txt,**/*.md"

Security tests with threshold

- uses: veriswarm/trust-check@v1
  with:
    api-key: ${{ secrets.VERISWARM_API_KEY }}
    agent-id: "agt_my_agent"
    mode: test
    fail-on-low-score: 70

Full check on deploy

- uses: veriswarm/trust-check@v1
  with:
    api-key: ${{ secrets.VERISWARM_API_KEY }}
    agent-id: "agt_production"
    action-type: deploy
    fail-on-deny: true
    min-trust-score: 70
    fail-on-low-score: 80

Outputs

Output	Description
trust-score	Composite trust score
policy-tier	Policy tier (allow/review/deny)
decision	Trust decision for the action
readiness-score	Security test readiness (0-100)
pii-found	Whether PII was found
injection-found	Whether injection was found
summary	Markdown summary of all checks