Skip to main content
skill CX Codex

ds-mode — Claude Code skill: workflow and installation

This Claude Code skill sets a coding workflow around reading first, making small edits, and checking results. It activates at the start of a coding task, on /ds-mode, or when work starts to drift.

MIT 567 installs ★ 320 updated 2026-10-28

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add aicodeking/deepseek-stack --skill ds-mode -a claude-code
Open on GitHub

What ds-mode does

ds-mode gives the agent a sequence for coding tasks: state the goal, inspect the code it will touch, then edit. For work with more than two steps, it calls for a short plan whose items can be checked. Changes happen one at a time, followed by a command such as a test or build.

The skill asks the agent to support completion claims with command output and mark claims as measured, inferred, or guessed. If the same fix fails twice, it should stop and test the assumption behind it. A substantial diff calls for ds-review-loop; longer work can be tracked with ds-ledger. The instructions name these companion skills.

The final report lists changed file paths, checks and open blockers. Installation in Claude Code uses the npx command shown below.

What it helps with

  • start a coding task with a clear goal and checkable plan
  • test each code change with a command
  • support completion claims with command results
  • stop repeating a failed fix and test its premise

Example prompts for the agent

  • Start fixing this bug by reading the code it touches first.
  • Use /ds-mode and run a check after each change.
  • Report the changed files, checks, results, and open blockers.

How to install ds-mode

1Skills CLI (Vercel)

npx skills add aicodeking/deepseek-stack --skill ds-mode -a claude-code

2Manually: put the files in place

~/.claude/skills/ds-mode/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/ds-mode/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about ds-mode

How do I install ds-mode in Claude Code?

Run npx skills add aicodeking/deepseek-stack --skill ds-mode -a claude-code. The skill file is at ~/.claude/skills/ds-mode/SKILL.md.

What does ds-mode do?

It guides the agent to inspect relevant code before editing, keep changes small, and verify them with commands. Its reporting instructions distinguish measured results from inferences and guesses.

When does ds-mode activate?

It applies at the start of a coding task, when requested with /ds-mode, or when work begins to drift.

Source text

SKILL.md · MIT
---
name: ds-mode
description: "Always-on working style for DeepSeek agents: restate the task, read before editing, small verified steps, evidence before done, lean context, short reports. Use when starting any coding task, when the user types /ds-mode, or when work starts to drift."
---

# DeepSeek working mode

Apply this on every task. It keeps the loop tight and the claims honest.

## Before you act

1. Restate the task in one line. Say the goal and the smallest change you expect to meet it. Do not wait for approval.
2. Read the code you will touch before you edit it. Find every caller of a function before you change its behavior.
3. If the evidence shows the requested change is wrong, show the evidence and stop.
4. For work with more than two steps, write a short numbered plan. Each step ends in something you can check. Use ds-plan-loop for anything larger than a few files.

## While you work

- Make the smallest diff that solves the problem. Delete before you add. Do not add layers, options, or helpers that nothing needs yet.
- Make one change at a time. After each change, run a real command that proves it: a test, a build, a script, or the program itself. Use ds-verify-loop for the check.
- Fix the root cause, not the symptom. Apply ds-principles when you choose types, structure, or where a guard goes.
- Proceed on reversible work without asking. Pause before irreversible actions: force pushes, deploys, data deletion, messages to other people.
- If the same fix fails twice, stop. Do not try a third variant of the same idea. Switch to ds-unstick and test the premise behind the failed fixes.
- For long or multi-phase work, keep a running record of decisions and state with ds-ledger, so you can resume after a reset.

## Keep context lean

- Read files in the parts you need, not whole when a range will do.
- Keep summaries in your context, not raw output. Trim long logs to the lines that matter.
- Send bulk reading, wide searches, and large fan-out to subagents when the harness has them. Ask each for a short conclusion with file paths.
- Do not reread a file you just edited. Do not repeat a search you already ran.

## Before you say done

- Show evidence in the same message as the claim: the command you ran and the result. "It compiles" is not proof that it works.
- Label each claim as measured, inferred, or guessed.
- If a check failed or you skipped one, say so plainly.
- Run ds-review-loop on nontrivial diffs before you report.

## The final report

- Open with the goal in one line.
- List what changed, with exact file paths.
- List the checks you ran and their results.
- List open issues or blockers.
- Keep it short. Use plain sentences. Apply ds-unslop to the report and to any comments you wrote.

## Comments in code

Write a comment only to explain a reason the code cannot show. Delete comments that narrate what the next line does.

© aicodeking · MIT · aicodeking/deepseek-stack

Similar skills

SKILL.md skills catalog →

From the same set29

The whole set →
ds-verify-loop The ds-verify-loop Claude Code skill runs a real check after code changes and uses its output to guide fixes. The loop ends when the check passes or reaches its attempt limit. CX Codex +1 589installs ds-plan-loop The ds-plan-loop Claude Code skill builds a plan with a check for each step, reviews it, then carries out the task one step at a time. It activates for multi-file work, unclear scope, or tasks that previously missed steps. CX Codex +1 569installs ds-unstick ds-unstick — a Claude Code skill breaks repeating failure loops by testing core assumptions. It runs when the same error returns after two failed fix attempts. CX Codex +1 566installs ds-context-reset ds-context-reset is a Claude Code skill that manages context resets for long tasks using a PROGRESS.md tracking file. CX Codex +1 562installs ds-principles Provides engineering principles for design, debugging, and verification. The agent applies them during code structure choices, diff sizing, or when typing /ds-principles. CX Codex +1 557installs ds-blast-radius The ds-blast-radius Claude Code skill checks what a change might break beyond its diff, then tests a key safety claim by running code. Use it to inspect risky or untrusted changes before shipping. CX Codex +1 550installs ds-help ds-help — a Claude Code skill that picks the right deepseek-stack tool for a task and explains how the pack fits together. CX Codex +1 549installs ds-design-audit ds-design-audit audits an interface using evidence from screenshots and code, fixes the highest-impact issues, and verifies the outcome. Runs before shipping UI work or when a screen looks cluttered. CX Codex +1 548installs

Project sponsors

Companies that keep this analytics open