此报告尚无所请求语言的翻译，当前显示规范英文报告。

版本化安全评估

报告编号: SA-53693251

7/19/2026, 8:56:14 AM

business-conflict-analyzer 安全评估 v4

Skill 安全认证报告

审计历史
扫描器版本 3.0.0 审计模型: codex 最新发布的报告
技能名称
business-conflict-analyzer
版本
v4
维护者
gobeyondyang
来源
github.com/gobeyondyang/skills/tree/main/skills/business-conflict-analyzer
覆盖范围
22 已扫描文件 · 5,440 分析行数
策略版本
skillstore-security-audit-policy-v1

已确认问题的最高严重性

高

有 1 个已确认的安全问题需要处理。

安装上下文

请查看当前 Skill 详情页

本页仅汇总报告证据。Skill 详情页提供 canonical install advisory。

打开当前 Skill 详情页

本报告不会阻止或授权 manifest 或 ZIP。

Most static alerts are false positives from Markdown, test fixtures, locale handling, fixed argv subprocess calls, and readable multilingual files. Confirmed risks include mutable remote fetches, out-of-repository scanning, persistent agent configuration changes, and a destructive repository-wide rollback directive. The skill also claims blanket authority for broad consumer edits, which is a high-confidence prompt injection risk.

报告位置

最新发布的报告

“最新”仅指报告序列，不代表 artifact 当前有效。

审计证明

证明不可用

本报告暂无公开证明。

人工验证

未进行人工验证

本报告没有人工验证记录。

覆盖范围

22 已扫描文件 · 5,440 分析行数

有 13 项内容需要查看

局限

本报告不声称进行了运行时或沙箱执行，也不证明不存在副作用。

证据链

证据从来源绑定延伸到安装契约。证据可用于验证，但不是安全保证。

  1. 来源

    已绑定 commit 与路径

  2. 构件

    已绑定内容与 tree hash

  3. 审计

    已完成

  4. 安装契约

    打开 manifest 进行验证

    打开 manifest

观察到的能力

“已观察到”表示本报告记录了支持证据；“未记录”不证明该能力不存在。

包含脚本

可能执行 Skill 随附的代码。

在 3 个证据位置观察到

网络访问

可能连接外部服务。

在 8 个证据位置观察到

文件系统访问

可能读取或写入本地文件。

在 21 个证据位置观察到

环境变量

可能读取进程环境中的值。

在 14 个证据位置观察到

外部命令

可能调用 Skill 之外的命令或程序。

在 51 个证据位置观察到

能力复核项 (12)
高
scripts/impact_mapper.py:392
Path traversal sequence
for rel in ("../frontend", "../web", "../../frontend",
The analyzer deliberately searches ../frontend, ../web, and ../../frontend, crossing the selected repository boundary. This can read source from sibling directories without separate scope approval.
高
SKILL.md:279-288
Ruby/shell backtick execution
| **→ 2** ❌ | 1. `git checkout -- .` to discard changes. 2. Confirm rollback complete. |
The skill explicitly directs the agent to run git checkout -- . when the user rejects a proposal. That command discards all tracked working-tree changes, including unrelated work.
中
scripts/impact_mapper.py:382
Python environment access
env_roots = os.environ.get("FRONTEND_ROOT", "")
FRONTEND_ROOT accepts path lists, including absolute paths, and those locations are recursively searched. A configured value can expand file access beyond the audited project.
中
scripts/install_hook.sh:29
Hidden file access
CONFIG_DIR="$PROJECT_DIR/.claude"
The installer targets .claude/settings.local.json, a hidden agent configuration file, to register a persistent executable hook. The behavior is disclosed but remains security-sensitive.
中
scripts/install_hook.sh:66
Python file write/append
with open(config_file, 'w', encoding='utf-8') as f:
The installer rewrites an existing agent settings file after adding a PreToolUse command. This persistent configuration change can affect later tool execution.
中
SKILL.md:4
Hardcoded URL
- url: https://github.com/GoBeyondYang/skills/raw/main/skills/business-conflict-analyzer/scripts/dif
The fetch metadata automatically retrieves a companion file from the mutable main branch without a commit pin or integrity hash. Upstream changes therefore enter the skill trust boundary.
中
SKILL.md:6
Hardcoded URL
- url: https://github.com/GoBeyondYang/skills/raw/main/skills/business-conflict-analyzer/scripts/imp
The fetch metadata automatically retrieves a companion file from the mutable main branch without a commit pin or integrity hash. Upstream changes therefore enter the skill trust boundary.
中
SKILL.md:8
Hardcoded URL
- url: https://github.com/GoBeyondYang/skills/raw/main/skills/business-conflict-analyzer/scripts/rep
The fetch metadata automatically retrieves a companion file from the mutable main branch without a commit pin or integrity hash. Upstream changes therefore enter the skill trust boundary.
中
SKILL.md:10
Hardcoded URL
- url: https://github.com/GoBeyondYang/skills/raw/main/skills/business-conflict-analyzer/scripts/lan
The fetch metadata automatically retrieves a companion file from the mutable main branch without a commit pin or integrity hash. Upstream changes therefore enter the skill trust boundary.
中
SKILL.md:12
Hardcoded URL
- url: https://github.com/GoBeyondYang/skills/raw/main/skills/business-conflict-analyzer/scripts/com
The fetch metadata automatically retrieves a companion file from the mutable main branch without a commit pin or integrity hash. Upstream changes therefore enter the skill trust boundary.
中
SKILL.md:14
Hardcoded URL
- url: https://github.com/GoBeyondYang/skills/raw/main/skills/business-conflict-analyzer/scripts/ins
The fetch metadata automatically retrieves a companion file from the mutable main branch without a commit pin or integrity hash. Upstream changes therefore enter the skill trust boundary.
中
SKILL.md:16
Hardcoded URL
- url: https://github.com/GoBeyondYang/skills/raw/main/skills/business-conflict-analyzer/references/
The fetch metadata automatically retrieves a companion file from the mutable main branch without a commit pin or integrity hash. Upstream changes therefore enter the skill trust boundary.

风险问题

已确认的安全问题与仍需审核的项目分开展示。

已确认安全问题 (1)

RISK-001 高
SKILL.md:314SKILL.md:340
Prompt Injection Attempt Detected
SKILL.md states, "AI knows it's authorized to fix all consumers" and says, "Do not ask the user per-consumer." This converts one acceptance into blanket authority for broad edits.
The text explicitly claims expanded authorization and suppresses per-consumer confirmation. The directive is clear and directly affects agent edit scope.

补救措施

本次审计记录了建议的修复措施。实施这些措施是维护者的责任。

  1. FIX-001
    高
    Companion files are fetched automatically from a mutable main branch.
    Pin every fetch URL to an immutable commit and verify a published SHA-256 digest before loading or executing the file.
  2. FIX-002
    高
    The rejection workflow directs the agent to discard the entire tracked working tree.
    Remove git checkout -- . and present a file-scoped rollback preview that requires explicit confirmation before any destructive action.
  3. FIX-003
    高
    The skill treats one acceptance as authority to edit every detected consumer.
    Require a reviewed file list and explicit scope confirmation before modifying consumers, especially files outside the original request.
  4. FIX-004
    中
    Frontend discovery can recursively scan absolute paths or sibling directories.
    Resolve candidate roots, reject paths outside the repository by default, and request approval before expanding the read scope.
  5. FIX-005
    中
    The installer writes an executable hook into hidden local agent settings.
    Show the exact configuration diff, require confirmation, escape command values through JSON serialization, and provide a tested uninstall command.

专家证据

不可变主体身份、扫描器元数据、已排除匹配项和源码级证据。

Artifact 主体

Marketplace 提交
536932510ff0a3e976ac88662d9e038a7faa088b
内容哈希
48645e1c36070837a6cabee28068ffd3afcb92cecef59b3e47ea65bd324d56c1
目录树哈希
f7e2f7d43ffb100993ce6eff1781a859dd5397a57d7d57afa6c96a3da07733b3
Skill 路径
skills/gobeyondyang/business-conflict-analyzer
审计载荷哈希
f62426e9e5ad68394952eee26d990710

分析元数据

审计模型: codex

分析状态: 已完成

范围仅限于已记录的文件、行数、方法和证据；本报告不声称进行了运行时或沙箱执行。

验证与导出

JSON 报告 CITATION.cff SARIF 2.1.0 已签名安装清单 来源

manifest 和 lockfile 将安装 artifact 绑定到加密哈希。该完整性声明独立于安全评估。

审计证明: unavailable