[security] Correct false docstring claim about command sanitization in LocalCommandLineCodeExecutor

[AUTHENSOR]

Summary

The LocalCommandLineCodeExecutor docstring (line 57) claims:

Command line code is sanitized using regular expression match against a list of dangerous commands in order to prevent self-destructive commands from being executed which may potentially affect the users environment.

This sanitization does not exist. Tracing the full execution path (_execute_code_dont_check_setup, lines 341-435), the only processing of LLM-generated code before execution is silence_pip (adding -qqq to pip install commands). No dangerous-command filtering, no regex matching, no blocklist.

Severity: MEDIUM
Class: False security claim creating a false sense of security

Impact

An operator reading the docstring would believe destructive commands are filtered. In reality, LLM-generated code (including rm -rf /, curl | sh, etc.) executes directly on the local machine with zero filtering.

Fix

Replace the false claim with a clear .. warning:: that states no sanitization is performed and recommends Docker-based executors for untrusted code.

Verification

• Traced full execution path: execute_code_blocks -> _execute_code_dont_check_setup -> file write -> asyncio.create_subprocess_exec
• Only processing: silence_pip(code, lang) at line 357
• No DANGEROUS, sanitize, block_list, or command filtering anywhere in the path
• Single-file docstring change

Dedup

Issues #7662 (WebSocket RCE) and #4873 (eval injection) are different surfaces. Nobody has reported the docstring/code mismatch specifically.

---

Generated by redthread. Single-purpose documentation fix.

[ErenAta16]

Confirmed independently: grepped the entire code_executors/local package for anything resembling command filtering (dangerous, blocklist, regex-match guards) before executing LLM-generated shell input, and the only hit in the whole file is the docstring sentence itself — no actual filtering exists anywhere in _execute_code_dont_check_setup. The fix correctly replaces the false claim with an explicit warning rather than trying to bolt on filtering as part of a docs PR, which is the right scope for this change.