dbs-benchmark Skill for Claude: Pick a Competitor Worth Copying and Copy It
Makes your assistant pick the competitor worth copying with five filters, then lay out how to copy it step by step.
- Installs
- 30.1K
- on skills.sh
- Stars
- 11K
- on GitHub
- Files
- 4
- besides SKILL.md
- Updated
- Oct 2026
- last repo commit
dbs-benchmark helps your assistant choose one benchmark that earns a lot, that you can understand and that you can copy. It runs candidates through five filters, then writes out the benchmark's chain of getting customers, converting, delivering and repeat sales, plus a copy path. The core rule is to copy what the benchmark does, detail by detail, and to drop talk about how you differ.
It triggers when you ask to find a benchmark, pick someone to imitate or analyze a competitor reference. With a search tool it looks up public information; without one it gives a first direction and asks you for candidate accounts, shops or links. The folder holds 5 files including SKILL.md and bundles no scripts. The README licenses dbskill under CC BY-NC 4.0, for non-commercial use.
Good for
- Founders choosing which competitor or peer to model
- Creators and sellers on content platforms comparing against top accounts
- Anyone who keeps saying 'but my case is different'
Install dbs-benchmark
Claude Code
- 1
Copy this prompt.
Install the dbs-benchmark skill in Claude Code. Run this command: claude plugin marketplace add dontbesilent2025/dbskill claude plugin install dbs-benchmark@dontbesilent-skills Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Claude Code.
The plugin route installs only this one skill. The README says the exact command name is whatever your agent shows; plugin installs may add a prefix. The README's whole toolkit command (claude plugin install dbs@dontbesilent-skills) is not needed for this skill.
From the project README. Source.
What's in the skill
When it loads
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。
Sections in SKILL.md
- 核心判断
- 1.0 到 1 阶段,先模仿
- 2.对标要求没有差异点
- 3.颗粒度决定模仿的质量
- 4.只看利润
- 5.降维模仿
- 6.主体差异是噪音
- 五重过滤
- A 组:评估对标对象
- B 组:检查用户的决策思路
- 工作流程
- Step 0:先读材料
- Step 1:选择路径
- 提问纪律
- 输出
- 默认形态
Key facts
- Installs on skills.sh
- 30,148 (Nov 2, 2026)
- GitHub stars (whole repo)
- 11,039 (Nov 2, 2026)
- Last repo commit
- Oct 28, 2026
- Language of SKILL.md
- Chinese; replies in the language you write in
- Intended use
- Personal study, research and non-commercial use; commercial use needs the author's authorization
- skills.sh audits
- Gen Agent Trust Hub pass, Socket pass, Snyk warn
What it includes
Copy first, match every detail, judge by profit only, ignore who you are
Copy first, match every detail, judge by profit only, aim at a benchmark earning about 10x your goal, ignore who you are
Six core rulesReads your words and attachments first, then restates the core question
Reads your words, links and screenshots, then notes your state, target income and any candidates
Step 0: read your materialRuns them through the five filters and gives a verdict
Path A: you have candidatesFinds more profitable peers; uses public information if a search tool exists
Finds more profitable peers; with a search tool it uses public information and names sources
Network ·
Path B: business, no candidatesGives the method and a candidate list ranked by profit and ease of copying
Gives the method and a candidate list ranked by profit and ease of copying; asks one question if no direction
Path C: not started yetCompares you with it item by item; each difference needs your reason or gets aligned
Path D: one benchmark chosenProfit about 10x your goal; estimates are labeled as guesses with their basis
Filter 1: does it make money?Can you write its customer chain as an SOP a new hire could follow?
Filter 2: can you understand it?Licenses, exclusive channels and capital limits count; skills you can learn do not
Filter 3: can you copy it?Only when your words show it: quotes you and asks which step you cannot do
Filter 4: leave yourself outIndustry outlook and red-ocean talk are not used to judge candidates
Filter 5: skip 'business essence'Health, caregiving or capital limits are recorded as real, and you decide
Real difficultiesVerdict, candidate table, profit chain, copy path, your noise table, items to verify
Output: default formThree to five lines when you ask about one candidate or one detail
Output: short formReads cases, platform-operations and templates notes when needed
Reads cases.md, platform-operations.md and templates.md when a case, platform or template is needed
Reference notesCopies approach, not whole texts; invents no revenue figures; drops illegal candidates
Copies approach, not whole texts or footage; invents no revenue figures; drops illegal or fraudulent candidates
LimitsVerdict first; offers "save" once a benchmark is locked; replies in your language
Verdict first; once a benchmark is locked it offers "save" (the README says /dbs-save does the saving); replies in your language
Style, save reminder, language
From the official docs
Permissions
Low risk
This skill works from what you type and from public information a search tool finds. Treat its earnings estimates as guesses.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
It may search the web for candidates
SKILL.md says it uses public information when a search tool exists and names the source. Without one it asks you for links. github.com
It estimates a competitor's earnings
The skill labels estimates as guesses with their basis and says it invents no revenue figures. Treat every number as a rough guess. github.com
Technical details
| Talks to | Public web sources, only if your assistant has a search tool |
|---|---|
| Scopes |
|
Questions
What does dbs-benchmark do?
It picks a business or account worth imitating, using five filters, and lays out how to copy it. It then compares your setup with the benchmark item by item.
How do I install dbs-benchmark?
Use the install box at the top of this page. For Claude Code the README documents a plugin that installs only this skill, and the skills CLI also works. Run install commands in a terminal, not in the agent chat.
Does it work in English?
Yes. The skill text is Chinese, but it says to reply in English when you write in English.
Does it copy someone's content?
Not whole texts. The skill says to copy approach, structure and visible details. A hook line may be reused word for word, but whole texts, footage, images and trademarks are not copied directly, and it warns about plagiarism risk.
Is dbs-benchmark safe to install?
The folder holds SKILL.md, an agent config file and three reference notes, and it bundles no scripts. SKILL.md says it uses public information when your assistant has a search tool. Share only what you are comfortable pasting into your assistant. Saving a conclusion is done by a separate skill, /dbs-save, which the README says writes to ~/.dbs/.
Can I use it for my company's paid work?
Check with the author first. The README and LICENSE say CC BY-NC 4.0, for personal study, research and non-commercial use. GitHub shows no detected license.
How is it different from dbs-diagnosis?
dbs-benchmark picks who to copy and how. The README describes dbs-diagnosis as judging whether a business problem holds up or checking a whole business.
Information
updated Nov 2, 2026
- Repository
- dontbesilent2025/dbskill
- Folder
skills/dbs-benchmark
Sources
Docs, issue trackers and discussions this page is built from.
Show all 6Hide
- dbskill README (English) · github.com
- dbs-benchmark on skills.sh · skills.sh
- dbs-benchmark SKILL.md · github.com
- dbs-benchmark templates.md · github.com
- dbs-benchmark cases.md · github.com
- dbs-benchmark platform-operations.md · github.com