Chat Commands - JoelBondoux/AtlasMind GitHub Wiki

Chat Commands

AtlasMind registers the @atlas chat participant in VS Code's built-in chat panel. Type @atlas followed by a slash command or a freeform question.

Slash Commands

Command Description
/bootstrap Initialize SSOT memory structure and optionally scaffold governance files
/import Scan an existing project and populate memory with metadata
/project Decompose a goal into subtasks, preview impact, and execute autonomously
/runs Open the Project Run Center to review recent autonomous runs
/agents List and manage registered agents
/skills List and manage registered skills
/memory Query the SSOT memory system
/cost Show session cost summary
/voice Open the Voice Panel for text-to-speech and speech-to-text
/vision Pick workspace images and ask a multimodal question

/bootstrap

Creates the SSOT memory folder structure and offers optional CI/CD governance scaffolding.

@atlas /bootstrap

What happens:

  1. Creates project_memory/ with all SSOT sub-folders
  2. Prompts for project type → populates project_soul.md
  3. Optionally scaffolds .github/workflows/ci.yml, PR template, issue templates, CODEOWNERS, .vscode/extensions.json
  4. Non-destructive — never overwrites existing files

/import

Scans the current workspace and auto-populates SSOT memory.

@atlas /import

What it scans:

  • Manifests: package.json, Cargo.toml, pyproject.toml, go.mod, pom.xml, build.gradle, Gemfile, composer.json
  • README files
  • Config: tsconfig.json, ESLint, Prettier, .editorconfig, Dockerfile, docker-compose.yml, Makefile, .gitignore
  • License files

What it creates:

  • architecture/project-overview.md — from README
  • architecture/dependencies.md — from manifest
  • architecture/project-structure.md — from directory listing
  • domain/conventions.md — from config files
  • domain/license.md — detected license type

Detected project types: VS Code Extension, API Server, Web App, Library, CLI Tool, Rust/Python/Go/Java/Ruby/PHP Project.

/project

Decomposes a goal into subtasks and executes them autonomously.

@atlas /project Refactor the auth module to use JWT tokens
@atlas /project Add comprehensive unit tests for the API layer
@atlas /project Set up a CI/CD pipeline with Docker deployment

Flow:

  1. LLM breaks the goal into a DAG of subtasks
  2. Preview shows estimated file impact
  3. If changes exceed the approval threshold (default: 12 files), you must approve
  4. Subtasks execute in parallel batches with ephemeral agents
  5. Final synthesis report streamed to chat
  6. Run saved to Project Run History

See Project Planner for full details.

/runs

Opens the Project Run Center to review, re-run, or inspect past autonomous runs.

@atlas /runs

/agents

Lists all registered agents with their roles and enabled status.

@atlas /agents

Output includes agent name, role, description, and whether the agent is currently enabled.

/skills

Lists all registered skills (built-in + custom + MCP) with their enabled status.

@atlas /skills

/memory

Queries the SSOT memory system by keyword.

@atlas /memory authentication decisions
@atlas /memory project architecture
@atlas /memory deployment runbooks

Returns matching entries ranked by relevance (title, path, tag, and snippet matches).

/cost

Shows the current session's cost summary.

@atlas /cost

Displays total cost in USD, total requests, and per-provider breakdown.

/voice

Opens the Voice Panel for text-to-speech (TTS) and speech-to-text (STT).

@atlas /voice

/vision

Opens an image picker for workspace images and submits a multimodal prompt.

@atlas /vision Describe what's in these screenshots

Freeform Chat

Any message without a slash command is treated as a freeform request:

@atlas How is error handling done in this codebase?
@atlas Write a function to parse CSV files with proper error handling
@atlas Explain the model routing algorithm

What happens behind the scenes:

  1. Orchestrator selects the most relevant agent
  2. Memory manager fetches related SSOT entries
  3. Task profiler infers phase, modality, and reasoning needs
  4. Model router picks the best model (within budget/speed preferences)
  5. Agent executes with available skills
  6. Response streamed to chat with cost tracking

Multimodal: Freeform messages auto-detect image paths in the workspace and attach them to the prompt.

Session context: The last N turns (configurable, default: 6) are carried forward for conversational continuity.

Extension Commands

These are also available from the Command Palette (Ctrl+Shift+P):

Command What it does
AtlasMind: Open Settings Budget/speed sliders, approval policies, verification config
AtlasMind: Manage Model Providers Add API keys, refresh models, health checks
AtlasMind: Manage Agents Create and configure custom agents
AtlasMind: Open Project Run Center Review, approve, pause, resume autonomous runs
AtlasMind: Manage MCP Servers Connect external tool servers
AtlasMind: Tool Webhooks Forward tool events to external endpoints
AtlasMind: Open Voice Panel TTS and STT
AtlasMind: Open Vision Panel Image-based multimodal prompts
AtlasMind: Bootstrap Project Same as /bootstrap
AtlasMind: Import Existing Project Same as /import
AtlasMind: Show Cost Summary Same as /cost
AtlasMind: Configure Scanner Rules View and edit skill security scanning rules

Follow-up Suggestions

After each command, AtlasMind suggests relevant next steps:

After Suggested follow-ups
/bootstrap View agents, View skills, Query memory, Start a project
/import View imported overview, View dependencies, View agents, Start a project
/project Review cost, Save plan to memory, Run another project
/agents View skills, Run a project, How to add an agent
/skills View agents, How to add a skill, Run a project
Freeform Turn into a project