fix(cron): catch ValueError on NUL paths in lifecycle_guard os.open

[rainbowgore]

What does this PR do?

Follow-up to #76762: _read_referenced_script() in cron/lifecycle_guard.py only caught OSError around os.open(path). Paths with an embedded NUL raise ValueError: embedded null byte (CPython), which escaped the guard, crashed the terminal tool, and surfaced as Failed to execute command: embedded null byte.

Widen the handler to except (OSError, ValueError) so the guard fails open (None, False) instead of crashing — same pattern already used at the Path.resolve() site from #76762.

Note: open PRs #78572 and #78013 also touch this area (broader NUL hardening). This change is the minimal #79398 fix only.

Related Issue

Fixes #79398

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• cron/lifecycle_guard.py — catch (OSError, ValueError) on os.open in _read_referenced_script
• tests/hermes_cli/test_gateway_restart_loop.py — add test_nul_bearing_script_path_does_not_crash_guard (issue repro)

How to Test

1. On unfixed main, run:

   from cron.lifecycle_guard import contains_gateway_lifecycle_command_or_referenced_script
   cmd = "source /tmp/hermes-snap-\x00x.sh >/dev/null 2>&1"
   contains_gateway_lifecycle_command_or_referenced_script(cmd, cwd="/Users/agents")

   Observe: ValueError: embedded null byte
2. On this branch, run the same snippet — expect False with no exception.
3. scripts/run_tests.sh tests/hermes_cli/test_gateway_restart_loop.py -q — all tests pass (including the new regression). Confirm hermes gateway restart is still detected as unsafe.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 26.5.1 (arm64)

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Screenshots / Logs

Before (on main):

UNCAUGHT CRASH -> ValueError embedded null byte
  File "cron/lifecycle_guard.py", line 260, in _read_referenced_script
    descriptor = os.open(path, flags)
ValueError: embedded null byte

After:

RESULT: False

Made with Cursor

[alt-glitch]

This was generated by AI during triage.

Duplicate of #77898 — both catch os.open ValueError for the same embedded-NUL lifecycle-guard path and preserve the existing fail-open result.

[CN-42]

补充一个真实生产环境复现案例（Linux / Ubuntu 26.04 LTS），确认此修复对实际触发场景有效：

真实触发场景

Hermes gateway 运行中（_HERMES_GATEWAY=1），agent 执行一条含二进制可执行文件路径的命令：

ls ~/.local/bin/cua-driver 2>/dev/null && ~/.local/bin/cua-driver --version 2>/dev/null | head -1

cua-driver 是 42MB 的 ELF 二进制（cua-driver 0.17.0）。_iter_referenced_shell_scripts() 把它当作"引用的脚本"（因为 token 含 /），_read_referenced_script() 调用 os.open(path) 直接崩溃。

完整 traceback（生产网关实测）

File "tools/terminal_tool.py", line 2560, in terminal_tool
    if contains_gateway_lifecycle_command_or_referenced_script(
File "cron/lifecycle_guard.py", line 353, in contains_gateway_lifecycle_command_or_referenced_script
    return _contains_unsafe_gateway_action(
File "cron/lifecycle_guard.py", line 335, in _contains_unsafe_gateway_action
    if script_text and _contains_unsafe_gateway_action(
File "cron/lifecycle_guard.py", line 324, in _contains_unsafe_gateway_action
    script_text, unsafe = _read_referenced_script(script_path)
File "cron/lifecycle_guard.py", line 260, in _read_referenced_script
    descriptor = os.open(path, flags)
ValueError: embedded null byte

表面现象：terminal 工具报 Failed to execute command: embedded null byte，整个命令未执行，会话表现为"卡住"。

与 #79398 / #79444 的差异

• [Bug] lifecycle_guard: os.open(path) crashes with ValueError: embedded null byte — incomplete fix for #76762 #79398 用 source /tmp/...\x00... 构造 NUL 路径；lifecycle_guard: unhandled ValueError ("embedded null byte") from os.open() aborts every terminal tool call #79444 用 cwd 含 NUL 触发
• 本案例是零构造的真实场景：命令本身完全正常，NUL 路径来自二进制文件内容被 _read_referenced_script() 读取后 tokenize 出的垃圾路径（cua-driver 是 42MB ELF，内容含大量 NUL），经 read_remote_script（cat 兜底）递归扫描时产生
• 说明任何普通命令只要引用了大型二进制可执行文件（cua-driver、firefox、python3 等）都可能触发，无需用户构造特殊输入

修复验证

应用本 PR 的 except (OSError, ValueError) 后（本地 patch 相同改动），原命令正常通过守卫（返回 False），不再崩溃。与 #79398 的 suggested fix 完全一致。

补充建议：_read_referenced_script() 的 NUL 防护已覆盖"文件内容含 NUL"（037825c1f）和"路径含 NUL"（本 PR），但 read_remote_script 的 cat 兜底路径对超大二进制（>1MB 被本地跳过后走 cat）仍会把整个二进制读进递归扫描——如果后续想更彻底，可在 _read_script_in_env 对二进制内容也做 NUL 检测。

[MattBetancourt]

Independent reproduction on Linux (Ubuntu 24.04, GCP e2-standard-2, Hermes v0.20.0/v2026.8.3).

Trigger: Any command referencing a path-like token that reaches _read_referenced_script with an embedded NUL — we hit it via the ELF-binary-content path (same traceback shape as the cua-driver report above: os.open on a path tokenized from binary data decoded as text).

Fix verified locally: Applied the identical except (OSError, ValueError) change at line 260-261. Added a regression test (test_null_byte_path_does_not_crash_guard) alongside the existing test_absolute_path_binary_does_not_crash_guard. All 83 tests in tests/hermes_cli/test_gateway_restart_loop.py pass.

# Repro (crashes on main, passes after fix):
from cron.lifecycle_guard import contains_gateway_lifecycle_command_or_referenced_script
contains_gateway_lifecycle_command_or_referenced_script(
    ".venv/bin/python\x00 -m pip install --quiet polymarket-us",
    cwd="/tmp",
)  # main: ValueError: embedded null byte → fix: False (no crash)

This is the minimal fix consistent with the Path.resolve() pattern already in place from #76762. PR has merge conflicts against current main — happy to help resolve if useful.