Scan types

ANTHRION runs nine scan types across six surfaces. Each one accepts a target (a URL, a public repository, or pasted text) and returns the normalized finding model. Start any scan from the dashboard at app.anthrion.xyz — see Your first scan for the full walkthrough.

AI / LLM attack scan#

The differentiator. An adaptive AI red-team: static probes escalate into an adaptive attacker that adjusts to your agent, rather than running a fixed checklist.

• Targets — a live AI endpoint (OpenAI-compatible), or a pasted system prompt to test it before you ship.
• Detects — prompt injection, jailbreaks, system-prompt leakage, excessive agency, insecure output handling, and related OWASP LLM Top-10 risks.
• Report — each finding names the technique, shows the exchange that demonstrated it, and recommends a mitigation.

Web app scan#

Dynamic testing (DAST) of a live website in a real browser.

• Targets — a public site URL.
• Detects — injection, cross-site scripting, authentication and misconfiguration issues on the rendered surface.

API security scan#

Probes an API for common server-side weaknesses.

• Targets — an API endpoint URL.
• Detects — broken authentication, injection and data-exposure issues across routes.

Web3 dApp scan#

Checks what a dApp's frontend could ask a wallet to do, plus on-chain context.

• Targets — a dApp URL.
• Detects — wallet-injection and frontend risks, and surfaces on-chain context for the contracts the dApp interacts with. Read-only — ANTHRION never signs or sends a transaction.

White-box scan (SAST)#

Static taint analysis of source code from a repository.

• Targets — a public GitHub repository URL.
• Detects — data-flow weaknesses such as SQL injection, command injection and path traversal, where untrusted input reaches a sensitive sink. Each finding shows the taint path.

Leaked-secrets scan#

Finds credentials committed to a codebase.

• Targets — a public GitHub repository URL.
• Detects — hardcoded passwords, API keys and high-entropy secrets. Disclosure is masked — the report shows enough to locate a secret, never the full value.

Code-similarity scan#

Measures how much a codebase overlaps with known code.

• Targets — a public GitHub repository URL.
• Detects — significant similarity to known libraries or other code, useful for spotting forks, copied code and license concerns.

GitHub trust score#

Aggregates repository health and red flags into a single trust signal.

• Targets — a public GitHub repository URL.
• Detects — a health score plus separate red flags drawn from repository metadata and the code/secret scans above.

x402 endpoint verify#

Checks whether a paid (x402) endpoint correctly and safely implements the protocol.

• Targets — an x402-enabled endpoint URL.
• Detects — a trust score (0–100) plus protocol red flags, so you — or an autonomous agent — can decide whether an endpoint is safe to pay before paying it. Read-only: the verify scan never pays.