Claude Agent SDK agent without a tool allowlist
MediumAV-FW-007ToolsFramework pack
What it finds
A Claude Agent SDK agent is created without allowed_tools, disallowed_tools, or a can_use_tool permission callback.
Why it matters
Without an allowlist the agent can use every built-in tool its permission mode allows, which can include running commands and editing files.
What an attacker could do: The agent may be able to run shell commands or modify files beyond what its task needs.
How to fix it
Pass allowed_tools with only the tools this task needs, and add a can_use_tool callback to approve or deny sensitive tool calls.
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
- CLAUDE_AGENT_NO_TOOL_ALLOWLIST
- How it detects
- Looks for a risky pattern, and stays quiet when a matching safeguard is present.
- Confidence
- Likely: strong signal, but review it in context.
- Scans
- Agent source using the named framework (LangChain, CrewAI, AutoGen, OpenAI Agents SDK, Claude Agent SDK, computer-use tools)
- OWASP LLM Top 10
- LLM08 - Excessive Agency
- NIST AI RMF
- GV-4.1, MS-2.3
- SOC 2
- CC6.1, CC6.3
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.