技能 biomni
📦
biomni
内容修订 r2 严重 ⚡ 包含脚本📁 文件系统访问🔑 环境变量🌐 网络访问⚙️ 外部命令
使用 Biomni 运行生物医学研究工作流
生物医学研究任务通常需要协调的数据检索、分析代码和模型配置。此技能指导 Claude、Codex 和 Claude Code 完成 Biomni 设置、任务模式、提供商选择和报告生成。
支持: Claude Codex Code(CC)
请勿自动安装此 Skill。
权威策略要求在执行任何安装操作前由操作人员审查。
使用我的智能体安装
将此请求复制给你的智能体，其中包含权威 Skill 页面和清单。
智能体请求
Review the Skillstore skill "biomni" from https://skillstore.io/skills/davila7-biomni.md and its manifest at https://skillstore.io/api/skills/davila7-biomni/manifest. Verify the artifact. Do not auto-install. Inspect the skill and report your findings, then wait for an operator or manual installation decision.你的智能体仍应展示其计划，并按安全策略请求必要确认。
Agent 可读资源
当 AI Agent、爬虫或脚本需要干净上下文，而不是读取完整页面时，请使用这些链接。
测试它
正在使用“biomni”。 一位研究人员要求为 HEK293 细胞中的自噬调节因子制定 CRISPR 筛选计划。
预期结果:
该技能应生成一个结构化工作流，包括候选通路、对照设计、sgRNA 选择标准、预期质量检查和审阅清单。
正在使用“biomni”。 一位平台工程师询问如何为 Biomni 配置 LLM 提供商。
预期结果:
该技能应比较提供商选项、所需凭据、模型选择、超时设置以及更安全的密钥存储实践。
正在使用“biomni”。 一位科学家要求基于 Biomni 对话历史生成可复现报告。
预期结果:
该技能应描述报告输入、支持的输出格式、需包含的元数据，以及共享生成文件前的安全检查。
安全审计
严重v9 • 2026/7/23 打开版本化报告
Most scanner hits are Markdown formatting, placeholder credentials, documentation links, or fixed subprocess calls. Confirmed issues include credential exposure, unsafe secret storage, generated-code injection, unescaped HTML, unpinned package execution, and full-privilege LLM code execution.
6
已扫描文件
2,860
分析行数
17
复核项
0
已忽略误报
已确认安全问题 (12)
严重
Unsandboxed LLM-Generated Code Execution
The skill states that Biomni executes LLM-generated code with full system privileges. Prompt or data manipulation could therefore become arbitrary host code execution.
The documentation explicitly states full-privilege generated-code execution and describes generation and execution as standard agent behavior.
高
Environment file access
# Or in .env file
The guide recommends storing an API key in a local .env file without requiring restrictive permissions or exclusion from version control.
高
Environment file access
echo "ANTHROPIC_API_KEY=sk-ant-..." >> .env
The shell command appends a credential to .env without setting file permissions. A substituted live key can become readable or accidentally committed.
高
Environment file access
env_file = Path.cwd() / '.env'
The path targets .env for credential storage. New files use ambient umask permissions, and the script applies no explicit owner-only mode.
高
Environment file access
# Read existing .env if present
The script reads every existing .env value and later rewrites the file alongside newly collected credentials.
高
Environment file access
# Write to .env
The following write stores API keys in plaintext .env content without restricting file permissions.
高
Environment file access
print("1. .env file (recommended, local to this directory)")
The setup recommends local .env storage but does not mention owner-only permissions or mandatory version-control exclusion.
高
Environment file access
Configure API keys (store in `.env` file or environment variables):
The instructions recommend .env credential storage without requiring restrictive permissions or version-control exclusion.
显示全部 12 条已确认问题
高
Environment file access
# Or check .env file in working directory
The troubleshooting guidance directs users to inspect a credential-bearing .env file, increasing accidental display or support-log disclosure risk.
高
Custom Path Enables Generated Python Injection
A user-supplied data path is interpolated without escaping into generated Python source. The setup script then executes that source.
A path containing quotes and newlines can break the Python string, insert statements, and reach the immediate subprocess execution.
高
scripts/generate_report.py:57-63scripts/generate_report.py:111-117scripts/generate_report.py:197-254
Stored Script Injection in HTML Reports
Conversation content and report titles enter HTML without escaping. A crafted conversation file can create active script or markup in the generated report.
The converter concatenates untrusted message text into tags and code blocks without HTML escaping or sanitization.
高
Unpinned MCP Package Executes with an API Key
The MCP example uses npx with automatic approval and no package version. The downloaded package receives the Brave API key and agent privileges.
The documented command executes the latest registry package through npx -y and provides a credential in its environment.
能力复核项 (17)
这些是真实的本地能力，对此技能可能是预期行为，因此需要复核，但不会按已确认恶意行为计分。
高
Python getenv function
print(os.getenv('ANTHROPIC_API_KEY'))
The troubleshooting example reads an API key and prints its value. Following it can expose a live secret in terminals, logs, or recordings.
高
getenv function call
print(os.getenv('ANTHROPIC_API_KEY'))
The getenv call is passed directly to print, disclosing the API key rather than checking only whether it exists.
高
Generic API/secret keys
print(os.getenv('ANTHROPIC_API_KEY'))
This example prints the full ANTHROPIC_API_KEY value. Terminal history, captured output, or support logs could retain the secret.
高
Generic API/secret keys
print(default_config.anthropic_api_key)
The example prints the configured Anthropic API key directly. It should report presence or a masked suffix instead.
高
Python subprocess.run
subprocess.run([sys.executable, str(test_file)], check=True)
The subprocess executes test_biomni.py after an unescaped custom data path is interpolated into that file. A crafted path can inject Python statements.
高
Python file write/append
with open(env_file, 'w') as f:
The script writes collected API keys to .env as plaintext without creating the file with owner-only permissions.
高
Python file write/append
with open(test_file, 'w') as f:
This writes Python source containing an unescaped user-selected path. The generated file is executed immediately on line 219.
高
Generic API/secret keys
anthropic_key = input(" Enter ANTHROPIC_API_KEY (or press Enter to skip): ").strip()
The script collects the Anthropic API key with input(), which echoes the secret on screen instead of using hidden credential input.
高
Generic API/secret keys
openai_key = input(" Enter OPENAI_API_KEY (or press Enter to skip): ").strip()
The script collects the OpenAI API key with visible input, exposing it to screen capture and nearby observers.
高
Generic API/secret keys
google_key = input(" Enter GOOGLE_API_KEY (or press Enter to skip): ").strip()
The script collects the Google API key with visible input rather than a non-echoing secret prompt.
高
Generic API/secret keys
groq_key = input(" Enter GROQ_API_KEY (or press Enter to skip): ").strip()
The script collects the Groq API key with visible input rather than a non-echoing secret prompt.
高
Generic API/secret keys
existing_vars.update(api_keys)
Collected API keys are merged into variables that are written to a plaintext .env file without an owner-only permission step.
高
Generic API/secret keys
for key, value in api_keys.items():
The loop sends each key value to print on line 159. Selecting shell export therefore discloses every collected credential on screen.
高
Generic API/secret keys
save_api_keys(api_keys, method='shell_export')
This branch calls shell_export, whose implementation prints each raw credential value to the terminal.
高
Generic API/secret keys
save_api_keys(api_keys, method='env_file')
This branch persists collected credentials in a plaintext .env file without enforcing owner-only permissions.
高
Generic API/secret keys
echo $ANTHROPIC_API_KEY
The troubleshooting command echoes the full Anthropic API key, which can expose it in terminal output or captured logs.
中
Python file write/append
with open(example_file, 'w') as f:
This writes an example Python file containing an unescaped custom path. Running the generated example can execute path-injected statements.
风险因素
⚡ 包含脚本 (1)
📁 文件系统访问 (6)
🔑 环境变量 (44)
references/api_reference.md:348 references/api_reference.md:355 references/llm_providers.md:363 references/llm_providers.md:363 references/llm_providers.md:218 references/llm_providers.md:219 references/llm_providers.md:27 references/llm_providers.md:30 references/llm_providers.md:56 references/llm_providers.md:74 references/llm_providers.md:96 references/llm_providers.md:113 references/llm_providers.md:125 references/llm_providers.md:149 references/llm_providers.md:168 references/llm_providers.md:184 references/llm_providers.md:203 references/llm_providers.md:260 references/llm_providers.md:363 references/llm_providers.md:367 scripts/setup_environment.py:82 scripts/setup_environment.py:88 scripts/setup_environment.py:93 scripts/setup_environment.py:95 scripts/setup_environment.py:100 scripts/setup_environment.py:102 scripts/setup_environment.py:107 scripts/setup_environment.py:109 scripts/setup_environment.py:114 scripts/setup_environment.py:116 scripts/setup_environment.py:118 scripts/setup_environment.py:122 scripts/setup_environment.py:125 scripts/setup_environment.py:142 scripts/setup_environment.py:158 scripts/setup_environment.py:294 scripts/setup_environment.py:296 scripts/setup_environment.py:304 scripts/setup_environment.py:306 scripts/setup_environment.py:316 scripts/setup_environment.py:327 scripts/setup_environment.py:328 SKILL.md:43 SKILL.md:289
🌐 网络访问 (20)
references/llm_providers.md:114 references/llm_providers.md:126 references/llm_providers.md:259 references/llm_providers.md:284 scripts/generate_report.py:189 scripts/setup_environment.py:41 scripts/setup_environment.py:92 scripts/setup_environment.py:99 scripts/setup_environment.py:106 scripts/setup_environment.py:113 scripts/setup_environment.py:343 scripts/setup_environment.py:344 scripts/setup_environment.py:344 SKILL.md:269 SKILL.md:270 SKILL.md:271 SKILL.md:272 SKILL.md:273 SKILL.md:307 SKILL.md:271
⚙️ 外部命令 (50)
scripts/generate_report.py:63 scripts/generate_report.py:69-71 scripts/generate_report.py:71-88 scripts/generate_report.py:88 scripts/setup_environment.py:25 scripts/setup_environment.py:47 scripts/setup_environment.py:67 scripts/setup_environment.py:219 SKILL.md:37-39 SKILL.md:39-41 SKILL.md:41-42 SKILL.md:42-45 SKILL.md:45-47 SKILL.md:47-51 SKILL.md:51-62 SKILL.md:62-70 SKILL.md:70-84 SKILL.md:84-87 SKILL.md:87 SKILL.md:88 SKILL.md:90-91 SKILL.md:91-94 SKILL.md:94-100 SKILL.md:100-120 SKILL.md:120-125 SKILL.md:125-133 SKILL.md:133-136 SKILL.md:136-145 SKILL.md:145-148 SKILL.md:148-159 SKILL.md:159-162 SKILL.md:162-171 SKILL.md:171-173 SKILL.md:173-184 SKILL.md:184-190 SKILL.md:190-198 SKILL.md:198-204 SKILL.md:204-218 SKILL.md:218-237 SKILL.md:237-238 SKILL.md:238-242 SKILL.md:242-251 SKILL.md:251-256 SKILL.md:256-258 SKILL.md:258-259 SKILL.md:259-260 SKILL.md:260-263 SKILL.md:263-265 SKILL.md:265-266 SKILL.md:266-280
审计者: codex 查看审计历史 →
分享与引用此报告
分享版本化评估报告、中性徽章、嵌入卡片和引用。Skillstore 只报告证据，不判断该 Skill 是否安全。
复制报告链接
https://skillstore.io/skills/davila7-biomni/audits/9?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown 徽章
[](https://skillstore.io/skills/davila7-biomni?utm_source=security_passport_badge)HTML 徽章
<a href="https://skillstore.io/skills/davila7-biomni?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/davila7-biomni/security.svg" alt="Skillstore security assessment" loading="lazy"></a>嵌入卡片
<iframe src="https://skillstore.io/embed/skills/davila7-biomni.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>学术引用 (APA · BibTeX · CFF)
APA 引用
davila7. (2026). biomni security audit report (audit version 9) [Author version unspecified]. Skillstore. https://skillstore.io/skills/davila7-biomni/audits/9BibTeX 引用
@techreport{davila7-davila7-biomni-2026,
author = {davila7},
title = {biomni security audit report (audit version 9)},
institution = {Skillstore},
year = {2026},
number = {9},
url = {https://skillstore.io/skills/davila7-biomni/audits/9},
note = {Author version unspecified}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "biomni security audit report (audit version 9)"
version: "unspecified"
type: report
authors:
- name: "davila7"
date-released: "2026-07-23"
url: "https://skillstore.io/skills/davila7-biomni/audits/9"
identifiers:
- type: other
value: "skillstore:davila7-biomni:audit:9"
description: "Skillstore immutable audit report identifier"
比较变体
2 个可安装变体每位作者的技能都保持为独立的可安装项。推荐变体按 Skillstore 证据排序。
为什么此变体排名第一
Skillstore 使用量更高
作者 / 来源 Skillstore 评分 证据置信度 Skillstore 使用量 已更新
Skillstore 评分
评分依据 证据置信度: 中 68
架构
85
可维护性
87
内容
71
社区
83
规范符合性
你能构建什么
设计基因组学筛选
使用清晰的输入、约束、预期输出和报告步骤来规划 CRISPR 或单细胞分析任务。
评估候选药物
为 Biomni 支持的药物发现工作流构建 ADMET、靶点发现和化合物优先级排序提示。
配置研究智能体
为共享研究环境设置提供商、数据路径、MCP 工具、超时和可复现的报告导出。
试试这些提示
启动基础基因查询
使用 Biomni 回答一个关于 [gene] 的聚焦问题。包括分子功能、已知疾病关联以及用于验证的关键参考文献。
规划 CRISPR 筛选
为研究 [cell type] 中 [phenotype] 的 CRISPR knockout 筛选创建一个 Biomni 任务。包括基因优先级排序标准、对照、预期分析输出和安全约束。
分析单细胞数据
指导 Biomni 分析 [dataset path]。执行质量控制、聚类、细胞类型注释、差异表达分析，并制定简明的解读计划。
设计多提供商研究运行
使用 [primary provider] 并在需要时回退到 [secondary provider]，为 [research question] 构建一个 Biomni 工作流。定义数据输入、超时限制、验证检查、报告输出和沙箱要求。
最佳实践
- 在启用自主代码执行前，在容器、虚拟机或受限环境中运行 Biomni。
- 使用最小权限 API 密钥，并按项目隔离生物医学数据集、凭据和生成的报告。
- 在每个研究提示中指定生物体、实验方法、数据集路径、预期输出和验证标准。
避免
- 不要在同时存储广泛云凭据或提供商凭据的主机上对敏感数据集运行 Biomni。
- 不要在未经专家审阅的情况下，将生成的生物医学发现视为已验证的临床或实验室决策。
- 在检查是否包含机密、患者数据和不安全的嵌入内容之前，不要共享生成的报告。
常见问题
Biomni 用于什么？
Biomni 支持 LLM 辅助的生物医学研究任务，例如基因组学分析、药物发现、临床基因组学和方案规划。
此技能是否包含 Biomni 包？
不。它提供指导和辅助脚本。用户仍需安装 Biomni 并配置所需的提供商凭据。
哪些 AI 工具可以使用此技能？
该报告列出支持 Claude、Codex 和 Claude Code。
Biomni 是否需要 API 密钥？
是。大多数工作流至少需要一个已配置的 LLM 提供商密钥，例如 Anthropic、OpenAI、Google、Groq、Azure 或 AWS Bedrock。
Biomni 对敏感生物医学数据安全吗？
需要谨慎。请在隔离环境中运行，限制凭据，审查生成的代码，并在处理敏感数据前确认数据处理规则。
这能否替代专家生物医学审阅？
不能。Biomni 输出在用于临床、监管或实验室决策前，应由合格专家审阅。
开发者详情
作者
davila7许可证
MIT
Skillstore 修订
r2
版本提示
作者未声明版本。
引用
c43861a65bb95efcae259cd161c9d6f4dc7eec6f
维护新鲜度
2026/7/24
使用量
16 次下载 · 268 次浏览
文件结构