Skip to content
SkillModel-led3,005 tokens in use

dbs-computer-use

One entry point for agent computer control on macOS, Windows and Linux: picks Cua Driver, agent-desktop or the built-in tool, works in the background and pauses before risky clicks.

11,518 stars on the repository Custom / Unknowndontbesilent2025/dbskill · skills/dbs-computer-use
npx -y skills add dontbesilent2025/dbskill -g --all
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing dbs-computer-use costs

202tokens idle
3,005tokens in use
8,114tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Fair
  • Frontmatter8/15
  • Structure16/25
  • Examples9/15
  • Output0/15
  • Size13/15
  • Resources12/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Excellent
  • Development activity96
  • Maintenance100
  • Community64
  • Maturity28
  • Momentum40

dbs-computer-use lets an agent operate web pages and desktop apps for you. It first asks whether the job needs the screen at all, since an API, a command line or a file is faster. If it does, it checks which control tools the agent has and uses the best one, preferring work in the background so your mouse stays yours.

When no tool is available it explains which one to install, what that enables and the official install command, and waits for your approval. Passwords, codes and payments are always left to you, and every key action is read back to confirm it worked.

What You Get

  • A tool choice based on the tools the agent has loaded, in the order Cua Driver, agent-desktop, then the agent’s own computer use
  • Install advice with the official command and what it enables, applied only after you agree
  • Background operation by element, with escalation to foreground only after you say yes
  • A pause before payments, deletions, sends, form submissions and account changes, naming the button and the effect
  • A report with the result, the tool used and anything left unconfirmed

Common Use Cases

  • Reading data from a logged-in web dashboard
  • Operating a desktop app that has no API
  • Setting up a computer control tool on a new machine
  • Checking a page without taking over your mouse

What Makes It Unique

Most computer-use setups assume one tool. This Skill detects capabilities by looking at the tool list, ranks three routes, and falls back one step at a time. It refuses to use browser extensions that act inside your signed-in browser because of account-ban risk. It treats a success message as unproven until the screen state is read back.

Under The Hood

The agent follows the steps in SKILL.md and reads reference files for each tool. It calls the computer-control tools your agent already has, such as Cua Driver, and runs the agent-desktop command line through the shell. It can fetch public web pages when only text is needed. When a tool is missing, it proposes shell install commands that download and run official scripts or use npm, and runs them only after you approve; macOS permission prompts are left for you to click. It can click, type and read in real apps and browsers. It does not enter passwords or make payments.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers