@blockrun/llm@3.24.4
⚠ Under reviewTypeScript SDK for BlockRun - every frontier model behind one OpenAI-compatible client. Pay with a BlockRun API key or per-call in USDC. Smart routing picks the cheapest capable model. No rate limits.
AI Security Review
scanned 2d ago · by lpm-firewall-aiThe inspected payment behavior is consistent with the package’s stated per-call payment function, and no concrete attack was found. The declared Anthropic SDK and viem dependency sources were unresolved in the supplied inspection, so the package cannot be marked clean.
Decision evidence
public snapshot- The manifest points to ESM and CommonJS entrypoints and lists no install lifecycle hook.
- The runtime signs the amount and recipient from an HTTP 402 payment challenge for USDC; this behavior is reached during a caller’s request.
- When no BlockRun wallet key resolves, the wallet helper creates and saves a new wallet.
- The manifest declares Anthropic SDK and viem dependencies; the supplied dependency inspection could not resolve their source, so their runtime behavior remains unreviewed.
- The inspected payment flow uses the service’s payment challenge and does not show a hidden transfer to a package-selected recipient.
- No concrete malicious behavior was established in the inspected package entrypoints.
Source & flagged code
5 flagged · loading sourceSource uses private key material to transfer cryptocurrency funds.
dist/index.jsView on unpkg · L101The runtime signs the amount and recipient from an HTTP 402 payment challenge for USDC; this behavior is reached during a caller’s request.
dist/index.jsView on unpkg · L5472Findings
Affected versions and remediation
This report applies to @blockrun/llm@3.24.4.
See all versions of @blockrun/llm for other recorded verdicts.
- Review the evidence and your use of @blockrun/llm@3.24.4 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.