AI Agent Risk Score
78/100
High exposure
Customer dashboard
License status, scans, reports, policies.
This is the post-purchase service hub customers use after buying a license.
CapitalGuard scan
acme-platform
78/100 High
AI Agent Risk Score
78/100
High exposure
Dangerous files
7
Require policy blocks
Exposed secret risks
4
Redacted in report
Unsafe permissions
5
Tool access review
Prompt-injection paths
3
Docs and issues
Policy file
1
Ready to deploy
Production secrets reachable.env.productionCritical
Prompt-injection path founddocs/prompts/release-checklist.mdHigh
Workflow edit risk.github/workflows/deploy.ymlHigh
Policy file generated.capitalguard/policy.ymlReady
Dangerous files
7
Require policy blocks
Exposed secret risks
4
Redacted in report
License timeline
Buyer workflow after checkout.
Payment verified
Stripe-hosted checkout records the package, buyer email, and campaign attribution.
License activation
Buyer activates the private key before private report and scanner downloads are released.
Authorized scope
Repository ownership or written authorization is required before any scan starts.
Report delivery
Findings are redacted, ranked, and packaged for founder, developer, and executive review.
Scan credits
A license has measurable inventory.
Included scans
10
Pro license allowance
Available scans
8
After two sample report slots
Policy exports
Unlimited
For the licensed organization
Update window
1 year
Included with fixed licenses
Delivery actions
Everything a buyer expects after paying.
Activate private key
Match purchase email and license key before private deliverables unlock.
ContinueSubmit authorized repo
Send scope, AI tool access details, and authorization evidence.
ContinueDownload buyer kit
Keep license terms, policy starter, safe-use guide, and firewall installer.
ContinueGenerate policy file
Turn findings into guarded AI-agent operating rules.
ContinuePolicy posture
Detection turns into precautions.
Secrets stay redacted in reports and screenshots.
High-impact workflow edits require human review.
Prompt files, docs, logs, and tickets are treated as untrusted input.
Monitoring can be added after the first licensed scan.