VibeCode Liability Index
Named failure modes of AI-built apps.
Named failure modes of AI-built apps, mapped to severity and fix order. Each entry describes the recurring launch risk, the surface it appears on, and the evidence VibeAudit can check without pretending unsupported areas are verified.
AI Coding-Agent Worm Static Indicators
Static repository evidence matched AI coding-agent worm patterns such as editor/agent auto-run config, secret-dump workflows, or GitHub/npm/cloud credential-harvesting code.
Read failure modeClient-Side Admin Fantasy
Admin or privileged flows appear to be enforced from client-side state or UI branches rather than server-side authorization.
Read failure modeCommand Injection Shell Game
User-controlled input appears to reach a shell execution sink without allowlist or argv-array evidence.
Read failure modeCommitted Secrets Exposure
Git history scanner evidence indicates that secret material was committed to repository history.
Read failure modeFrontend Key Confessional
A browser-facing file or public environment surface appears capable of exposing private tokens, service keys, or operational secrets.
Read failure modeIDOR Object Ownership Gap
Resource identifiers appear to drive object access without nearby owner or tenant predicate evidence.
Read failure modeKnown Malicious Dependency
A dependency version matched a known malicious, compromised, or campaign-flagged package in the supplied public advisory feeds.
Read failure modeSQL Injection Slipstream
User-controlled input appears to reach raw or dynamically composed SQL execution without parameterization evidence.
Read failure modeWeak Password Hashing
Password creation or verification code appears to use a fast general-purpose hash instead of a password storage KDF.
Read failure modeAuth Middleware Missing Person
Auth dependencies or routes are visible, but middleware or server-side gate proof is not obvious in the digest.
Read failure modeKnown Vulnerable Dependency
A dependency version matched a known vulnerability advisory in the supplied public advisory feeds.
Read failure modePath Traversal File Escape
User-controlled paths or filenames appear to reach filesystem access without base-directory containment evidence.
Read failure modeSSRF Open Proxy
URL-like user input appears to reach a server-side HTTP request without host/protocol allowlist evidence.
Read failure modeSupabase RLS Roulette
Supabase or database access is present without enough proof of row-level security, policy tests, or private server mediation.
Read failure modeXSS Output Escape Hatch
Raw HTML rendering appears without sanitizer evidence.
Read failure modeDatabase Door Left Open
Database clients or schema files are present without enough visible proof of least-privilege access or migration discipline.
Read failure modeStripe Cosplay
Payment code exists, but the repository does not prove a complete checkout, webhook verification, and entitlement path.
Read failure modeWebhook Trust Fall
Webhook paths appear present without visible signature verification or idempotency proof.
Read failure modeAgent With a Weapon
Agent or tool-execution paths are present without enough proof of sandboxing, scoped permissions, and prompt-injection boundaries.
Read failure modePrompt Injection Welcome Mat
LLM input, retrieval, or tool paths are present without visible prompt-injection tests or policy boundaries.
Read failure modeFile Upload Roulette
File upload or storage flows are present without enough visible proof of size, type, malware, or path controls.
Read failure modePublic Bucket Jump Scare
Storage/bucket code or naming suggests public object exposure may exist without proof of private access controls.
Read failure modeDependency Time Bomb
Dependencies are present without enough proof of lockfile hygiene, audit status, or update discipline.
Read failure modeCORS Free-For-All
API surfaces exist without visible proof of restrictive CORS, security headers, or production header policy.
Read failure modeRate Limit Missing Person
Public API or auth endpoints are present without visible rate-limit or abuse-control proof.
Read failure modePublic Source Map Confessional
Source map artifacts or production source map settings may expose implementation details after deployment.
Read failure modeSecurity Header Paper Shield
Deployment header configuration lacks one or more baseline browser security headers.
Read failure modeRun Quick Scan to see the public diagnosis layer for a checked external scope. Private findings and remediation order stay gated to paid reports.
Run Quick Scan