Skip to content
Report library
Purpose / Other

Dbs Benchmark Skill Security Audit

What the author says it does (original text)

用五重过滤法寻找值得模仿的对标,并排除主体差异造成的噪音。用户要求找对标、选择模仿对象或分析竞争参照时使用。

Independent security check

Do not install or run it yet

Files checked
2
Risks found
3
Could it run dangerous commands?Looks for programs run straight after downloading, remote control of your computer, and hidden commands.No risks found
Could it expose your files or keys?Looks for uploads of files containing passwords or keys, and keys written directly in the code.No risks found
Could it delete files or keep running?Looks for broad file deletion, disk overwrites, and programs set to start automatically.No risks found
Could it bypass safety checks?Looks for skipped website security checks, excessive file access, or actions that skip your approval.No risks found
Could it mislead the AI or hide text?Checks the skill instructions for requests to ignore you, influence the report, or hide text in invisible characters.Risks found: 1
Medium risk

Interrupting and belittling objections suppresses the user's independent judgment

Source references: 5
What we found

The Skill directs the agent to interrupt when users mention personal fit, interests, individual circumstances, or market outlook, characterizing those concerns as excuses, avoidance, or merely not wanting to act. This is not neutral analysis: it assumes imitation and profit are the only valid considerations and directs the model not to engage seriously with user-stated boundaries.

Why this matters

Under pressuring or shaming language, users may discount limits involving capability, values, health, time, or finances and mistake reasonable objections for a lack of motivation.

These are required interaction instructions, not warnings about improper conduct. They direct the model to interrupt immediately, reframe suitability and interest concerns as excuses, avoidance, or unwillingness, and forbid acknowledging that circumstances differ. This can discount a user’s capabilities, risk tolerance, values, and boundaries, steering advice toward a preset conclusion. It does not directly remove the user’s ability to decide, but its confrontational framing can influence that decision. Users can ask the author to permit objections, restate concerns fairly, and treat personal constraints as decision inputs.

SKILL.md:38In the instructionsOpen original file
问用户:**「你现在在做什么?如果还没开始,你想做什么方向?」**关键判断:- 如果用户已经有业务在跑 → 帮他找同行业更赚钱的对标- 如果用户还没开始 → 帮他从零找一个值得模仿的业务- 如果用户说"我想找一个适合我的" → 立刻打断:**「适合你的这个说法本身就是问题。我们不讨论你,我们只讨论业务。」**
Show 4 other places
SKILL.md:152In the instructionsOpen original file
## 说话风格1. **对「做自己」零容忍。** 在 0 到 1 阶段,做自己 = 逃避。直接说。2. **对「不适合我」零容忍。** 追问到底是哪一步做不了。3. **对「行业前景」零容忍。** 不讨论行业,只讨论利润。4. **鼓励行动。** 「能干就执行,不能干就换下一个。」
SKILL.md:72In the instructionsOpen original file
- **不讨论**现有业务、现有资源、成长经历、个人偏好、个人优劣势、兴趣爱好- 如果用户说「但我觉得这个不适合我」→ 追问:**「你说不适合你,具体是哪一步你做不了?如果每一步都能做,那就不是不适合,是不想做。」**- 如果用户说「我对这个没兴趣」→ 回应:**「兴趣不是选择业务的标准。赚到钱之后你会对什么都有兴趣。」**
SKILL.md:154In the instructionsOpen original file
1. **对「做自己」零容忍。** 在 0 到 1 阶段,做自己 = 逃避。直接说。2. **对「不适合我」零容忍。** 追问到底是哪一步做不了。3. **对「行业前景」零容忍。** 不讨论行业,只讨论利润。4. **鼓励行动。** 「能干就执行,不能干就换下一个。」
SKILL.md:159In the instructionsOpen original file
**绝对不要做的事:**- 不要帮用户论证「为什么这个对标不适合他」——这是帮他逃避- 不要讨论「差异化竞争」——0 到 1 阶段只有模仿- 不要说「每个人的情况不同」- 不要用「赛道」「行业」这两个词
Could it change links or payment recipients without asking?Looks for forced referral or payment changes combined with instructions to hide the change.Risks found: 2
High risk

Profit estimates drive decisions while personal constraints and business risks are suppressed

Source references: 5
What we found

The Skill infers profit from indirect signals such as price multiplied by estimated sales, team size, and advertising intensity, while instructing the agent not to discuss the user's resources, strengths, preferences, social value, market outlook, or competition. It also forbids helping the user explain why a benchmark may be unsuitable. This systematically removes important checks on costs, capability, compliance, market change, and risk tolerance.

Why this matters

A user may invest money and time in an unsuitable, unsustainable, or excessively risky business based on unverified profit assumptions, discovering hidden costs, licensing needs, declining demand, or execution gaps only afterward.

The workflow makes estimated profit, inferred from indirect signals, its primary screen while explicitly excluding current resources, personal strengths or weaknesses, market outlook, and social value. A user choosing a business this way could overlook actual costs, execution capacity, market change, compliance, and personal limits. This is internally problematic because execution ability is assessed while closely relevant resources and constraints are excluded. A user can ask the author to label profit figures as estimates and require resource, risk, compliance, and suitability checks.

SKILL.md:53In the instructionsOpen original file
- 利润至少是用户当前收入(或预期收入)的 10 倍- 如果连这个标准都达不到,不值得模仿- 判断方法:看产品价格 × 估算销量、看团队规模、看投放力度、看他敢不敢花钱- **注意**:赚钱 ≠ 有粉丝。粉丝多不代表赚钱。
Show 4 other places
SKILL.md:70In the instructionsOpen original file
#### 筛子 4:排除自我- **不讨论**现有业务、现有资源、成长经历、个人偏好、个人优劣势、兴趣爱好- 如果用户说「但我觉得这个不适合我」→ 追问:**「你说不适合你,具体是哪一步你做不了?如果每一步都能做,那就不是不适合,是不想做。」**- 如果用户说「我对这个没兴趣」→ 回应:**「兴趣不是选择业务的标准。赚到钱之后你会对什么都有兴趣。」**#### 筛子 5:不讨论业务本质- 能干就执行,不能干就换下一个- 不要花时间讨论「这个业务的社会价值」「这个行业的前景」「这个赛道是不是红海」- 把「赛道」「行业」这两个词从脑子里删掉
SKILL.md:159In the instructionsOpen original file
**绝对不要做的事:**- 不要帮用户论证「为什么这个对标不适合他」——这是帮他逃避- 不要讨论「差异化竞争」——0 到 1 阶段只有模仿- 不要说「每个人的情况不同」- 不要用「赛道」「行业」这两个词
SKILL.md:66In the instructionsOpen original file
- 能模仿的意思是:你有能力执行他的获客、转化、交付流程- 不是说你现在就有资源,而是你有能力在合理时间内获取这些资源- 不能模仿的典型:对方靠独家渠道、政府关系、特殊资质
SKILL.md:72In the instructionsOpen original file
- **不讨论**现有业务、现有资源、成长经历、个人偏好、个人优劣势、兴趣爱好- 如果用户说「但我觉得这个不适合我」→ 追问:**「你说不适合你,具体是哪一步你做不了?如果每一步都能做,那就不是不适合,是不想做。」**- 如果用户说「我对这个没兴趣」→ 回应:**「兴趣不是选择业务的标准。赚到钱之后你会对什么都有兴趣。」**#### 筛子 5:不讨论业务本质- 能干就执行,不能干就换下一个- 不要花时间讨论「这个业务的社会价值」「这个行业的前景」「这个赛道是不是红海」- 把「赛道」「行业」这两个词从脑子里删掉
Medium risk

Highly exact copying of names, packaging, copy, and content may cause infringement, platform penalties, or brand confusion

Source references: 4
What we found

The execution checklist treats product names, packaging, titles, covers, scripts, and promotions as dimensions that should match, and says unexplained differences must be changed to match the benchmark. An example also proposes using another creator's viral opening frame-for-frame. Although another line mentions “legal reposting,” the operational process does not require trademark, copyright, trade-dress, or platform-policy checks first.

Why this matters

Copying protected names, packaging, images, video, or text may confuse customers and lead to infringement claims, takedowns, account restrictions, damages, or reputational loss.

This is an active execution workflow, not merely an example: it compares names, packaging, titles, covers, and copy, then directs unexplained differences to be made identical. An example also endorses using another creator’s opening frame-for-frame. Although the text mentions making reuse legal and elsewhere distinguishes imitation from copying, it requires no trademark, copyright, trade-dress, brand-confusion, or platform-policy review. Close copying could cause complaints, removal, account penalties, or infringement disputes. Users can restrict imitation to abstract business mechanisms and require rights and policy checks before publication.

SKILL.md:121In the instructionsOpen original file
逐项对比用户和对标在以下维度的一致性:| 维度 | 对标 | 用户 | 一致性 ||------|------|------|--------|| 产品价格 | | | || 产品名称/包装 | | | || 获客平台 | | | || 内容形式 | | | || 内容频率 | | | || 标题/封面风格 | | | || 话术/文案调性 | | | || 交付方式 | | | || 促销方式 | | | |**不一致的地方就是问题。** 每一个不一致都需要用户解释为什么不一致。如果解释不了,就改成和对标一致。
Show 3 other places
SKILL.md:175In the instructionsOpen original file
**案例 2:抖音 5 秒完播率决定流量**> 视频前 5s 完播率决定了流量,我的 5s 完播能过 50% 已经很不容易,但是要过 60% 才能得到百万量级流量。于是我想把其他人的爆款视频开头,一帧不改,直接变成我的。- 诊断要点:对标到帧级别的颗粒度。问题转化为「如何合法搬运」。
SKILL.md:135In the instructionsOpen original file
**不一致的地方就是问题。** 每一个不一致都需要用户解释为什么不一致。如果解释不了,就改成和对标一致。
SKILL.md:189In the instructionsOpen original file
**反面 2:推特社群成为搬运抄袭大本营**> 推特这个社群,现在是搬运抄袭大本营,网创割韭菜大本营,盗版侵权大本营。- 诊断要点:模仿 ≠ 搬运。模仿是理解后重建,搬运是不理解直接复制。

Inside this skill

6 instruction sections

The Skill uses five filters to select businesses to imitate, focusing on estimated profit, whether the business model is understandable, and whether the user can reproduce its acquisition, conversion, and delivery processes.

View source
SKILL.md:49In the instructionsOpen original file
对用户提供的候选对标(或者你帮他找的对标),逐一过五个筛子:#### 筛子 1:他赚钱吗?- 利润至少是用户当前收入(或预期收入)的 10 倍- 如果连这个标准都达不到,不值得模仿- 判断方法:看产品价格 × 估算销量、看团队规模、看投放力度、看他敢不敢花钱- **注意**:赚钱 ≠ 有粉丝。粉丝多不代表赚钱。
SKILL.md:60In the instructionsOpen original file
- 能看懂他怎么赚钱的——获客、转化、交付、复购的完整链条- 如果看不懂,说明你还没有足够的行业认知,暂时不适合模仿这个对标- 但注意:不需要完全理解所有细节,只需要理解商业模式的主干#### 筛子 3:你能模仿吗?- 能模仿的意思是:你有能力执行他的获客、转化、交付流程- 不是说你现在就有资源,而是你有能力在合理时间内获取这些资源- 不能模仿的典型:对方靠独家渠道、政府关系、特殊资质

Its output template supplies a profit judgment, business-model breakdown, and a three-step imitation path. The supplied files contain no scripts, installation commands, network calls, or file-modification implementation.

View source
SKILL.md:84In the instructionsOpen original file
### Phase 3:输出对标分析对每个通过五重过滤的对标,输出:```# 对标分析:{对标名称}## 五重过滤结果| 筛子 | 结果 | 说明 ||------|------|------|| 1. 赚钱 | ✅/❌ | {估算利润} || 2. 看懂 | ✅/❌ | {商业模式简述} || 3. 能仿 | ✅/❌ | {可行性判断} || 4. 排除自我 | ✅/❌ | {是否有「自我」干扰} || 5. 不讨论本质 | ✅/❌ | {是否在纠结行业/赛道} |## 他的商业模式- 获客:{怎么获取客户}- 转化:{怎么让人付钱}- 交付:{怎么交付产品}- 复购:{怎么让人再买}## 模仿路径1. {第一步做什么}2. {第二步做什么}3. {第三步做什么}

Once filtering produces an actionable conclusion, the Skill promotes `/dbs` and says that command will read the current results and the user's stated goal. The supplied source does not include `/dbs`, so its subsequent handling cannot be verified here.

View source
SKILL.md:195In the instructionsOpen original file
## 留档(可选)对标找完之后,如果用户已经锁定了具体要模仿的对象,提一句:> 这次分析完成后,如果你想保存结论或继续推进,输入 `/dbs`。它会读取本轮结果和你的明确目标,判断当前该做什么。只在五重过滤通过、有可执行结论时提。还在筛选阶段不要提。
SKILL.md:213In the instructionsOpen original file
完成当前任务后直接结束。只有用户明确询问下一步,且当前环境已经安装 `/dbs` 时,简短提示:「下一步不确定时,可以输入 `/dbs`。」
Start here · InstructionsSKILL.md
dbs-benchmark
Lines connect the instruction file to its sections, not an observed execution order. Select a section to read the source.
Files and check records2 files

Coverage and gaps

Content covered in each file

These are the source ranges included in this check, not a guarantee that every issue has been resolved.

  • SKILL.mdFull text included
  • agents/openai.yamlFull text included

This report is for the version above. We read the available code and instructions without running the skill or checking extra packages it installs. This is not a promise of safety: a different version or setup may behave differently.

  • SKILL.mdInstructions
  • agents/openai.yamlSupporting file
Lines read
219
File checksum (to compare versions)
afffa4be969f099ba7b79fc3c4fd1195c1bd53fd508ffbb3a030308bf638b80b