Evidence-led source reviews
Source traces follow lower-trust inputs through a failed security boundary to a concrete result. Only confirmed records receive severity; hardening suggestions are separate.
CAPABILITIES & LIMITATIONS
Useful findings. Honest boundaries.
An AI-assisted review supports your security programme. It does not guarantee secure code, replace every penetration test, or provide compliance certification.
Three distinct review profiles
Quick is a bounded partial pass. Standard adds separate candidate and final-record verification. Deep adds lifecycle detail and independent repeat coverage. All profiles preserve the same evidence bar.
Explicit coverage gaps
Budget exhaustion, failed model calls, missing infrastructure, cancelled jobs, and unverified candidates are retained as incomplete workflow state. They never become a “no vulnerabilities” claim.
Execution is conditional
Target-controlled code can run only with an independently configured isolation service meeting network, filesystem, resource, environment, and artifact-promotion controls. Missing controls block execution.
Integration availability
GitHub is the initial repository provider. GitLab, Bitbucket, private runners, and Kenyan payment methods are future extensions and are not currently advertised as working integrations.