Security context
What an agent needs to avoid regressing past fixes and find the next vuln in this repo.
pchalasani/claude-code-tools
main @ 3299c65
11
Fixes
0
CVEs
HIGH
Peak severity
0.0%
Coverage
Highlights
Auth Bypass: 9 prior fixes. Scrutinize any change in this area.
claude_code_tools/agent_tunnel/backends.py: most-fixed (2 issues). Treat as high-risk during review.
4 high-severity fixes in this history; regressions here are high-impact.
Recurring patterns
The bug types that recur here, drawn from past fixes, not open vulnerabilities.
Auth Bypass: Command validation routines were repeatedly bypassed via nested subshell execution ($()) and chained command operators because of non-greedy parsing. Secure command sanitization requires recursive extraction and parsing of all nested subcommands.
Auth Bypass: Handle-name collisions (such as the sentinel 'cli' handle) from different sessions could escalate agent-tunnel thread privileges, and arbitrary command execution was possible without explicit double opt-in checks.
Auth Bypass: Flawed JSON output schemas caused safety-hooks denial decisions to be ignored entirely, and missing tool validation allowed direct reads of sensitive local .env files containing secrets.