Back to catalog / Agent safety
Distinguish tool auto-approval from tool availability
Verify that allowed_tools pre-approves listed tools and use disallowed_tools to block explicit exclusions.
Goal
Avoid confusing “auto-approved” with “only these tools are available” when configuring tool permissions.
Steps
- Set
allowed_tools=[...]inClaudeAgentOptions, listing only tools that may be auto-approved. [readme] - Use
disallowed_tools=[...]for tools that must be blocked; other unlisted tools may still be decided by permission_mode or can_use_tool. [readme] - Verify allow/deny behavior with a read-only probe in a test directory, then enable write tools one at a time. [readme]
How to verify
Confirm blocked tools are explicitly in disallowed_tools and unapproved tools follow the expected permission path.
Troubleshooting
If a tool remains callable, check whether it was merely omitted from allowed_tools; omission does not disable it.
Permissions
Prerequisites
- Python 3.10+, claude-agent-sdk; validate in an isolated test directory.
Permissions
- Start with read-only tools; authorize any write or command execution separately.
Limitations
CLI configuration and permission_mode also affect tools; this list alone is not a filesystem or network sandbox.
Pinned upstream source
- Repository
- anthropics/claude-agent-sdk-python
- File path
- README.md
- Pinned commit
b6e9d12fe1cc98dde988ab7b7713c1feeee50c6c- Source checked
- 2026-10-14 · Reviewed against the pinned upstream source
Repository metrics
These values describe the upstream repository, not popularity or likes for an individual skill or playbook.
- Stars
- Not collected
- Forks
- Not collected
- Observed at
- Not collected
No metrics snapshot is available yet.
License: MIT
Main topic
Make Agent behavior safer and easier to verify