dbskill Claude skills

A Chinese business-diagnosis skill toolbox of 35 skills, led by a dispatcher that routes problems to execution, pricing or AI-writing-detection skills.

Our verdict Recommended with caveats

Thoughtfully designed with real safety guardrails, but the numbered-prompt feature fetches content from a remote GitHub repository at runtime and the license forbids commercial use.

What dbskill does

dbskill is a set of Chinese-language skills aimed at diagnosing business and personal-productivity problems rather than writing code. The entry skill, dbs, is a dispatcher: it reads the user's situation, decides whether to act directly, ask a clarifying question, combine several skills, or run a numbered hidden-tier prompt fetched from the project's public GitHub repository, verified by SHA-256 against a signed manifest before use. dbs-action diagnoses why someone knows what to do but is not doing it, using an Adlerian psychology frame, explicitly refusing to give generic encouragement and stopping instead of counseling if self-harm language appears. dbs-ai-check scans text for signs of AI-generated writing and, by design, only reports findings by default; it argues that removing AI traits is not the same as making writing better, and only offers rewrite help after the user explicitly asks and answers follow-up questions about intent. The entry skill also includes a self-update check that calls the project's own GitHub-hosted UPDATE.json over HTTPS. Across 35 skills the content averages about 12.8 KB with reference files and a handful of scripts, and the repo is actively maintained with CI.

Good for

  • Chinese-speaking founders and freelancers diagnosing pricing, execution or content problems
  • Writers who want AI-writing-pattern detection without an automatic rewrite
  • Users who want a single dispatcher skill rather than remembering many commands

Not for

  • English-only users, since content and prompts are in Chinese
  • Anyone seeking motivational coaching, which the execution skill explicitly refuses to give

What's inside depth 3/5

35 skills averaging 12.8 KB with reference files and a handful of scripts including a signed remote-fetch and verification mechanism, but no tests or evals.

Skills
35
Reference docs
61
Scripts
29
Tests / evals
3
Templates
11
Avg SKILL.md
12.5 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Python 3 and curl for the update-check and numbered-prompt scripts
  • Network access to raw.githubusercontent.com for update checks and the hidden-tier prompt library
  • Telemetry: The dispatcher skill checks a version file at raw.githubusercontent.com/dontbesilent2025/dbskill/main/UPDATE.json over HTTPS with a 5-second timeout on each use; this is a version check, not usage analytics, and failure is silently tolerated.

Security review

Broad access

This skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.

  • Loads instructions from a server at run timeWhen the user enters a three-digit code, /dbs fetches catalog.json and numbered-prompts/<code>/PROMPT.md from the repo's main branch on raw.githubusercontent.com at runtime and the skill executes that text as a workflow (SKILL.md line 55); the sha256 check uses a catalog fetched from the same place, so it does not pin content.skills/dbs/scripts/numbered-prompts.py:19

A large Chinese-language business/content skill pack with Python/Node/shell helpers. It checks GitHub for updates, pulls hidden numbered prompts from the repo's main branch at runtime and follows them, can run an unpinned npx skills add update when you reply 1, and symlinks itself into several agent skill folders. Video extraction needs your own TikHub/Qingdou key and sends the video links you supply to those services.

Worth knowing

  • Makes network requestsEach use of /dbs (at most once per 24h) fetches UPDATE.json from raw.githubusercontent.com and prints a notice; dbs-video-extract also calls TikHub (api.tikhub.dev, mcp.tikhub.io) and Qingdou (qingdou.vip) with the video links you give it.skills/dbs/scripts/check-update.sh:27
  • Needs an API key for its own serviceTikHub/Qingdou keys are read from env vars, a key file, ~/.config/dbs/API_Keys.md or a macOS Keychain entry named by the skill and sent as Bearer tokens only to those services (base URL overridable via TIKHUB_BASE_URL / QINGDOU_BASE_URL).skills/dbs-video-extract/scripts/tikhub_api.py:201
  • Runs remote code at an unpinned versionUpdating runs npx -y skills add dontbesilent2025/dbskill -g --all (unpinned, latest main) when you ask or reply 1 to the update notice; README also uses npx skills.skills/dbs-update/SKILL.md:39
  • Changes files outside your projectCreates/removes symlinks or Windows junctions for skills in ~/.agents/skills and in existing ~/.claude, ~/.hermes, ~/.kiro and other agent skill folders (and a Grok bridge); configure_api_key.py writes the pasted key in plaintext to ~/.config/dbs/API_Keys.md (mode 0600); state is kept in ~/.dbs.skills/dbs-install-skill/scripts/install-skill.sh:221
  • Relies on a CLI or package we could not reviewShells out (no shell) to lark-cli/feishu-cli for Feishu doc subscriptions and, with --allow-agent, to a user-configured agent_command; other skills call ffmpeg, swift, the claude plugin CLI and gh.skills/dbs-human-dispatch/scripts/dispatch.py:262

What we checked

The repository (259 files) on Oct 29, 2026. Not a full review: 1 of 259 files were not downloaded (large repository; we reviewed the skill folders); we read 4 of 58 code files; it relies on code outside the repository: Remote prompt files fetched at runtime from github.com/dontbesilent2025/dbskill main (skills/dbs/numbered-prompts/*); TikHub and Qingdou web APIs, and the lark-cli / feishu-cli, ffmpeg, swift and `skills` (npx) tools; knowledge base file 知识库/原子库/atoms.jsonl is missing from our copy.

NVIDIA SkillSpector static scan: 327 pattern hits (Agent Snooping 189, Prompt Injection 59, MCP Rug Pull 33, Dangerous Code Execution 24, Data Flow 11, Rogue Agent 4). The two CRITICAL tainted-flow hits are the TikHub and Qingdou clients sending the user's own service key to those services. Prompt-injection hits are zero-width characters and a 'Do Anything Now' mention inside an included book (epub/markdown) of tweets, not instructions to the agent; sudo/apt lines are in the project's CI workflow; Skill Enumeration hits are README links.

A review, not a certification. Skills change after we check them — read the files before installing.

License

CC BY-NC 4.0 — personal, research and non-commercial use with attribution only; commercial use needs the author's authorization.

Can it run unattended?

Not really. It is a conversational diagnostic tool that works through back-and-forth with the user, not an unattended task runner.

Alternatives to dbskill

Humanize KR (im-not-ai)

epoko77-ai/im-not-ai

A Korean-language humanizer that detects and rewrites AI-sounding text (translationese, mechanical parallelism) while preserving facts, with deterministic gates.

Recommended Check first Depth 4/5

anti-slop

dmmulroy/anti-slop

Opinionated vendored Oxlint rules that reject low-evidence TypeScript and JavaScript patterns, installed and configured by a bundled setup skill.

With caveats Check first Depth 3/5

FAQ

What is dbskill?

dbskill is a Chinese-language toolbox of 35 skills for diagnosing business and personal-productivity problems, such as pricing, procrastination and whether writing reads as AI-generated, dispatched through one entry skill.

How do I install dbskill?

Run npx -y skills add dontbesilent2025/dbskill -g --all for everything, or use the Claude Code plugin marketplace (claude plugin marketplace add dontbesilent2025/dbskill) to install individual skills such as dbs or dbs-diagnosis.

Can I use dbskill for commercial work?

No, not without the author's permission. It is licensed CC BY-NC 4.0, which allows personal, research and non-commercial use with attribution only.

Is dbskill safe to use?

The dispatcher fetches a version file and, for numbered prompts, signed content from the author's own GitHub repository, verified by checksum before use. We found no hidden payloads, but it does make network calls to a third-party-controlled repo.

Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026). Review based on the README and SKILL.md files as published.