技能 dittobot
📦
dittobot
内容修订 r1 严重 ⚙️ 外部命令🌐 网络访问📁 文件系统访问🔑 环境变量⚡ 包含脚本
用你自己的声音重写散文
通用 AI 编辑常常会抹平个人声音，并加入缺乏依据的润色。Dittobot 会重写用户提供的文本，同时保留事实、立场、节奏和约束。
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 "dittobot" from https://skillstore.io/skills/regionallyfamous-dittobot.md and its manifest at https://skillstore.io/api/skills/regionallyfamous-dittobot/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“dittobot”。 一条发布说明太长且不清楚，但必须提到 importer 修复和重试支持。
预期结果:
一条简洁的发布说明，保留 importer 修复、重试支持和平静语气，同时不添加新主张。
正在使用“dittobot”。 一篇帖子听起来泛泛而夸大，但源文本没有包含支撑细节。
预期结果:
一个更朴素的版本，删除空泛主张，并添加一条简短说明，询问真实细节。
正在使用“dittobot”。 一个个人段落中有一个奇特意象，承载了作者的声音。
预期结果:
一个更干净的段落，保留该意象不变，只修正其周围的铺陈。
安全审计
严重v5 • 2026/7/9 打开版本化报告
The strongest confirmed issue is the documented curl-to-bash installer path, which executes remote mutable content without checksum verification. The install tooling also modifies the hidden user skills directory, and optional live evaluation can send prompts and an API key to an external endpoint. Most other static findings are false positives from Markdown examples, GPL text, regression fixtures, or defensive redaction code.
33
已扫描文件
7,114
分析行数
23
复核项
0
已忽略误报
已确认安全问题 (2)
严重
Pipe to shell pattern
curl -fsSL https://raw.githubusercontent.com/RegionallyFamous/dittobot/main/install.sh | bash
The README explicitly documents curl piped directly into bash. This is a critical supply-chain pattern because remote content executes without local review or checksum verification.
高
Remote Installer Executes Environment-Selected Archive
install.sh lets DITTOBOT_REPO, DITTOBOT_REF, or DITTOBOT_ARCHIVE_URL choose the archive source, then downloads it and runs scripts/install.py from the unpacked result. This creates supply-chain risk if the environment or archive source is attacker-controlled.
The control flow is direct: environment variables can choose the archive URL, curl downloads it, tar unpacks it, and python3 runs scripts/install.py from the result. Legitimate installer customization is possible, but the environment-selected source makes this a high-confidence supply-chain risk.
能力复核项 (23)
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。
严重
Hardcoded URL
curl -fsSL https://raw.githubusercontent.com/RegionallyFamous/dittobot/main/install.sh | bash
The README URL is part of a curl-to-bash installation command. It asks users to fetch remote shell content and execute it immediately.
高
Hidden file in home directory
default=os.path.expanduser("~/.agents/skills/dittobot"),
The installer defaults to ~/.agents/skills/dittobot and can create or replace files in a hidden user skills directory. This is intended install behavior but still a real filesystem modification.
高
Python environment access
api_key = os.environ.get("OPENAI_API_KEY")
live_eval.py reads OPENAI_API_KEY from the environment. That is a legitimate opt-in secret, but the script handles a credential.
高
Generic API/secret keys
api_key: str,
The API key is passed as a function parameter for the outbound Responses API call. This is normal API usage but still a secret-bearing code path.
高
Generic API/secret keys
"Authorization": f"Bearer {api_key}",
The API key is placed in an Authorization bearer header. That is required for the API call, but it is a direct credential transmission path.
高
Generic API/secret keys
api_key = os.environ.get("OPENAI_API_KEY")
live_eval.py reads OPENAI_API_KEY from the environment before calling the API. This is opt-in, but it confirms secret access.
高
Generic API/secret keys
api_key=api_key,
The environment-sourced API key is passed into call_responses_api. That creates a credential-bearing path to network request code.
中
Hardcoded URL
ARCHIVE_URL="https://github.com/${REPO}/archive/${REF}.tar.gz"
install.sh builds a GitHub archive URL that is later downloaded and unpacked. This is real network egress used in the installer, even though the default host is GitHub.
中
Hardcoded URL
ARCHIVE_URL="https://github.com/${REPO}/archive/refs/tags/${REF}.tar.gz"
install.sh builds a GitHub tag archive URL that is later downloaded and unpacked. This is legitimate installer behavior, but it depends on remote content.
中
Hardcoded URL
ARCHIVE_URL="https://github.com/${REPO}/archive/refs/heads/${REF}.tar.gz"
install.sh builds a branch archive URL that is later downloaded and unpacked. Using a mutable branch for installation creates supply-chain exposure.
中
Hidden file access
default=os.path.expanduser("~/.agents/skills/dittobot"),
The installer expands a hidden home-directory target for the skill install. This is expected, but it touches user configuration state outside the repository.
中
Python shutil operations
shutil.copy2(source, destination)
install.py copies packaged files into the target install directory. The operation is scoped, but running the installer modifies the user filesystem.
中
Python shutil operations
shutil.rmtree(temp_target)
install.py removes an existing temporary install path before copying files. The path is derived from the target install path, so this is a real deletion operation.
中
Python shutil operations
shutil.move(str(target), str(backup))
install.py moves an existing target install to a backup path. That can alter an existing skill installation if the script is run.
中
Python shutil operations
shutil.move(str(temp_target), str(target))
install.py moves the completed temporary copy into the hidden skill directory. This is intended but changes user agent configuration state.
中
Python shutil operations
shutil.rmtree(temp_target)
install.py removes the temporary target during cleanup. The deletion is scoped, but it is still a filesystem operation by installer code.
中
Python HTTP libraries
import urllib.request
live_eval.py imports urllib.request for optional live model evaluation. The script can perform outbound API calls when invoked with an API key.
中
Python HTTP libraries
request = urllib.request.Request(
live_eval.py constructs an HTTP request containing the skill text and selected fixture prompt. This is opt-in, but it sends local evaluation data to an API endpoint.
中
Python HTTP libraries
with urllib.request.urlopen(request, timeout=timeout) as response:
live_eval.py opens the configured API URL with urllib.request.urlopen. This is real network egress during optional live evaluation.
中
Hardcoded URL
DEFAULT_API_URL = "https://api.openai.com/v1/responses"
The default endpoint is https://api.openai.com/v1/responses. It is a legitimate default, but it confirms that live_eval.py can contact an external service.
中
Python file write/append
return os.fdopen(fd, "a", encoding="utf-8")
live_eval.py can append transcript records to a local JSONL file. The file may contain model outputs or sources when raw save options are enabled.
中
Python os file operations
os.chmod(target, 0o600)
live_eval.py changes transcript file permissions to 0600. This is protective, but it confirms filesystem writes by the live evaluation tool.
低
Python environment access
parser.add_argument("--api-url", default=os.environ.get("OPENAI_API_URL", DEFAULT_API_URL))
OPENAI_API_URL can configure the outbound evaluation endpoint. The script requires HTTPS and an explicit allow flag for custom URLs, so the risk is mitigated but real.
风险因素
⚙️ 外部命令 (26)
install.sh:28 install.sh:50 LICENSE:317-319 LICENSE:321 LICENSE:323 scripts/build_plugin.py:90 scripts/check_install.py:88 scripts/failure_fixture.py:130-132 scripts/install.py:97 scripts/redact_case.py:117 scripts/redact_case.py:212-214 scripts/regression_100.py:176 scripts/regression_100.py:405-1291 scripts/regression_100.py:1291 scripts/scorecard.py:48 SKILL.md:47 SKILL.md:76-80 SKILL.md:80-119 SKILL.md:119-120 SKILL.md:120-121 SKILL.md:121 SKILL.md:122 SKILL.md:132-134 SKILL.md:134-136 SKILL.md:136 SKILL.md:138
🌐 网络访问 (15)
install.sh:10 install.sh:12 install.sh:14 README.md:187 README.md:193 scripts/build_plugin.py:41 scripts/build_plugin.py:43 scripts/build_plugin.py:44 scripts/build_plugin.py:63 scripts/build_plugin.py:64 scripts/build_plugin.py:65 scripts/live_eval.py:19 scripts/live_eval.py:134 scripts/live_eval.py:143 scripts/live_eval.py:32
📁 文件系统访问 (17)
install.sh:51 install.sh:18 install.sh:28 scripts/build_plugin.py:80 scripts/build_plugin.py:159 scripts/check_install.py:44 scripts/check_install.py:44 scripts/check_install.py:89 scripts/install.py:55 scripts/install.py:55 scripts/install.py:48 scripts/install.py:78 scripts/install.py:83 scripts/install.py:88 scripts/install.py:95 scripts/live_eval.py:73 scripts/live_eval.py:72
🔑 环境变量 (15)
scripts/live_eval.py:444 scripts/live_eval.py:445 scripts/live_eval.py:530 scripts/live_eval.py:117 scripts/live_eval.py:138 scripts/live_eval.py:242 scripts/live_eval.py:496 scripts/live_eval.py:530 scripts/live_eval.py:531 scripts/live_eval.py:532 scripts/live_eval.py:560 scripts/redact_case.py:25 scripts/redact_case.py:23 scripts/redact_case.py:24 SECURITY.md:27
⚡ 包含脚本 (1)
检测到的模式
Pipe to shell pattern
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/regionallyfamous-dittobot/audits/5?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown 徽章
[](https://skillstore.io/skills/regionallyfamous-dittobot?utm_source=security_passport_badge)HTML 徽章
<a href="https://skillstore.io/skills/regionallyfamous-dittobot?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/regionallyfamous-dittobot/security.svg" alt="Skillstore security assessment" loading="lazy"></a>嵌入卡片
<iframe src="https://skillstore.io/embed/skills/regionallyfamous-dittobot.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
RegionallyFamous. (2026). dittobot security audit report (audit version 5) [Author version unspecified]. Skillstore. https://skillstore.io/skills/regionallyfamous-dittobot/audits/5BibTeX 引用
@techreport{regionallyfamous-regionallyfamous-dittobot-2026,
author = {RegionallyFamous},
title = {dittobot security audit report (audit version 5)},
institution = {Skillstore},
year = {2026},
number = {5},
url = {https://skillstore.io/skills/regionallyfamous-dittobot/audits/5},
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: "dittobot security audit report (audit version 5)"
version: "unspecified"
type: report
authors:
- name: "RegionallyFamous"
date-released: "2026-07-09"
url: "https://skillstore.io/skills/regionallyfamous-dittobot/audits/5"
identifiers:
- type: other
value: "skillstore:regionallyfamous-dittobot:audit:5"
description: "Skillstore immutable audit report identifier"
Skillstore 评分
评分依据 证据置信度: 中 91
架构
85
可维护性
87
内容
67
社区
91
规范符合性
你能构建什么
润色工作消息
将粗略的邮件、更新和公告改写成清晰的信息，同时保留原始事实和立场。
保护个人声音
改进帖子、简介、说明文字和文章，同时不磨掉幽默、节奏或有用的粗粝边缘。
验证重写质量
使用内置本地工具测试事实保留、声音忠实度、约束和适合公开使用的夹具。
试试这些提示
整理草稿
Use $dittobot on this. Keep my facts and voice, but make it clearer and shorter: [paste text]
创建语气选项
Use $dittobot on this. Give me three versions: warmer, sharper, and concise. Preserve the claims: [paste text]
只诊断不重写
Use $dittobot to diagnose this draft. Do not rewrite it. Give the three highest-impact notes: [paste text]
构建声音档案
Use $dittobot to infer a reusable voice profile from these samples. Include use, avoid, rhythm, protected quirks, and when not to apply: [paste samples]
最佳实践
- 粘贴源文本，并说明受众、目的和任何硬性约束。
- 标出必须在编辑后保留下来的事实、姓名、日期、主张和短语。
- 当语气带有主观性或消息存在社交风险时，请要求提供多个选项。
避免
- 当声音保留很重要时，不要在没有源材料的情况下要求写一份新草稿。
- 不要把声音档案视为将旧事实导入新作品的许可。
- 不要为普通编辑任务运行维护者脚本。
常见问题
Dittobot 是做什么的？
它会重写和编辑所提供的散文，同时保留作者的事实、声音、立场和约束。
它会从零开始写作吗？
它可以帮助处理已有声音风格下的草稿，但它是为用户提供的文本而设计的。
它能处理敏感或高风险写作吗？
它会保留不确定性和精确性，但不能替代法律、医疗、财务或事实审查。
正常使用需要网络访问吗？
不需要。技能说明是本地的，但可选的实时评估脚本可以调用 OpenAI API。
什么是声音档案？
它们是根据样本推断出的紧凑编辑规则。它们用于指导风格，而不会导入旧事实。
哪些工具可以使用它？
报告列出了对 Claude、Codex 和 Claude Code 的支持。