Vercel AI SDK agent allowed hundreds of steps
MediumAV-FW-010Runtime safetyFramework packNew in 2.0
What it finds
A Vercel AI SDK call allows 100 or more tool steps per request.
Why it matters
Each step is a model call plus a tool call. A loop or an injection can run up large bills and many side effects before stopping.
What an attacker could do: One request can trigger hundreds of tool calls.
How to fix it
Keep the step limit to what the task needs (often 5–10) and add a per-user budget.
Pro shows the corrected code for your exact file, step by step, and re-scans to confirm the fix. See plans
Check details
- Finding code
- VERCEL_AI_HIGH_MAX_STEPS
- How it detects
- Looks for a specific risky pattern in the code or config.
- Confidence
- Definite: the risky setting is written plainly in the file.
- Scans
- Agent source using the named framework (LangChain, CrewAI, AutoGen, OpenAI Agents SDK, Claude Agent SDK, computer-use tools)
- OWASP LLM Top 10
- LLM04 - Model Denial of Service, LLM08 - Excessive Agency
- NIST AI RMF
- MG-2.2
- SOC 2
- A1.1
Check your agent for this
Paste your agent code or MCP config. The free scan runs all 124 checks in about 10 seconds — no account needed.
Related checks
- Command injection riskAV-RUNTIME-004 · Critical
- Unsafe dynamic code executionAV-RUNTIME-005 · Critical
- Privileged or host-mounted container configurationAV-RUNTIME-007 · Critical
- SQL query built from unescaped inputAV-RUNTIME-009 · Critical
- Unsafe deserialization of untrusted dataAV-RUNTIME-010 · Critical
- Prompt injection surface detectedAV-RUNTIME-002 · High