BitssyBytes logoAgent Safety Score

Directory / Claude Code

A

Claude Code

Coding agent (CLI) · Anthropic · 7/7 verified · Last checked 2026-10-16

Deny-first permission rules, PreToolUse hooks that can block even in bypass mode, and an OS-level Bash sandbox with domain allowlists. The sandbox only covers shell commands; file tools and MCP servers sit outside it.

Overall: A (2.57 avg, 7/7 controls verified)

Lock this down first

Turn on the Bash sandbox with strictAllowlist, and never use --dangerously-skip-permissions outside a container or VM.

Scorecard

CriterionRatingNotes
Permission granularity3/3Deny-first rules. A PreToolUse hook that denies blocks the tool even in bypassPermissions mode (hooks guide).
Human approval gates3/3Ask rules and permission modes decide what prompts; deny and ask rules hold even when a hook says allow.
Sandboxing2/3OS sandbox covers Bash only. Docs recommend a container or VM for unattended runs.
Secrets handling2/3Keychain on macOS, 0600 file on Linux. apiKeyHelper for custom credential scripts.
Audit logs2/3OpenTelemetry metrics and log events, opt-in. ConfigChange hooks can audit settings changes.
Kill switch / interrupt3/3Esc stops the current tool call, Ctrl+C interrupts, Ctrl+X Ctrl+K stops background subagents.
Network egress control3/3allowedDomains plus strictAllowlist for sandboxed commands.

Sources

Vendor? Request a correction. We review corrections within 7 days.