agent-tools

The skill/document serves as a safety-conscious reference for configuring tool permissions in Claude Code agents. There is no evident capability to exfiltrate data, install unverifiable binaries, or perform autonomous harmful actions. The footprint is coherent with a governance/configuration guide. The only elevated risk pattern is the documented option to bypass permissions in extreme cases, which should be used with explicit user intent. Overall, classify as BENIGN with low security risk.