Skip to content
Cursos/DeepSeek Harness (DSH)/DeepSeek Harness · Phase 4
Revisión pendiente desde 21 ago 20262 fuentes primarias

Execution modes: choosing the right one

A mode or profile determines the agent's tools and permissions. Give it the least power required for the mission.

Qué vas a conseguir ahora

Una decisión o prueba aplicada a «Execution modes: choosing the right one».

  1. 1Entiende el criterio
  2. 2Haz una práctica pequeña
  3. 3Guarda una evidencia
Evidencia

Una nota breve con qué hiciste, qué salió bien, qué falló y qué revisarías después.

Objetivos de aprendizaje
  • Separate inspection, editing and automation.
  • Create conservative profiles for each task.
  • Prevent an agent from doing more than requested.

dsh entry modes

The CLI README describes `dsh` as a launcher for profiles. A profile is an ordered stack of plugin-bundle patch layers under your own overrides.

The key commands are web, headless and plugin. Web opens the interface; headless runs one persisted task and exits; plugin forwards arguments to pnpm in the profile directory.

Terminal
dsh web
# alias for --profile web

dsh --profile headless "Summarize this repo in 5 bullets"

dsh plugin --profile my-profile add plugin-package

dsh --help

Profile before impulse

Specific profile names may change with the product, but the pattern is stable: one for inspection, one for controlled changes and one for experimentation. Do not use a full shell profile to summarize a folder.

  • Read-only: search and files, no writing.
  • Bounded change: workspace edits plus explicit tests.
  • Automation: only after repeatable proof and approvals for external actions.

Launcher arguments and profile arguments

The launcher parses its own flags first and passes the rest to the booted profile. Order matters: `--profile` belongs to the launcher; `--port` or `--resume` may belong to the profile app.

Terminal
dsh --profile web --port 8080
# --port belongs to the Web UI

dsh --profile headless "run the tests"

dsh --profile web --help
# web app help

dsh --help
# launcher help

Prompts that set boundaries

Prompt constraints do not replace permissions, but they reduce ambiguity and make the intended boundary auditable.

Terminal
Analyze the issue and propose a five-step plan.
Do not write files or run commands until I approve the plan.
State risks, affected files and verification steps.
Comprueba que funciona. For one real task, write a permission profile and a prompt with an explicit boundary.
Guardar y reabrir el proyecto.
The right mode shrinks the blast radius of an error.

Si has guardado la evidencia de esta lección, continúa con «Workspaces, permissions and safety». Si no, repite la comprobación antes de avanzar.

Complete Aulafy mapSee how this lesson fits without leaving your path.

Complete Aulafy map

How all courses connect

This is not a checklist. Start with the foundation, choose an outcome, and go deeper only when your project needs more control.

  1. 1Understand
  2. 2Apply or build
  3. 3Operate with confidence
01

Choose an application

Turn the foundation into a visible outcome: a website, a business improvement, media, or an interactive experience.

Continue into the technical branch when you need to maintain code, data, or infrastructure.

02

Build with code

Prepare your environment, work with coding agents, and run models while keeping control of your projects.

This branch prepares you to design and operate reliable AI systems.

03

Take systems to production

Combine retrieval, agents, evaluation, security, deployment, and model adaptation when the problem requires it.

You do not need every course: choose the component your system needs and return as it grows.

View full catalogue