Agentic coding workflow

For developers

The desktop obsidian CLI exposes developer commands for testing and debugging plugins and themes through the running app. These capabilities do not exist in the documented Headless ob surface.

Loop

sequenceDiagram
    participant Agent
    participant CLI as obsidian CLI
    participant App as Obsidian app
    Agent->>CLI: plugin:reload id=my-plugin
    CLI->>App: reload plugin
    App-->>CLI: ok
    Agent->>CLI: dev:errors
    CLI->>App: dump JS errors
    App-->>Agent: error buffer
    Agent->>CLI: dev:screenshot path=after.png
    CLI->>App: capture screenshot
    App-->>Agent: base64 PNG written to path
    Agent->>CLI: eval code="app.vault.getFiles().length"
    CLI->>App: run JS in context
    App-->>Agent: result

Useful commands

Step	Command
Reload a plugin you are developing	obsidian plugin:reload id=my-plugin
Open dev tools	obsidian devtools
Attach the Chrome DevTools Protocol debugger	obsidian dev:debug on
Capture a screenshot	obsidian dev:screenshot path=screenshot.png
Read the JS error buffer	obsidian dev:errors
Read the console buffer	obsidian dev:console limit=50 level=error
Inspect CSS for a selector	obsidian dev:css selector=".workspace-leaf"
Query the DOM	obsidian dev:dom selector=".cm-line" text
Run arbitrary JS	obsidian eval code="app.workspace.getActiveFile()?.path"

Tips

• Pair dev:errors and dev:console with a plugin:reload cycle so the agent only sees diagnostics from the reload it just triggered.
• The eval command runs JavaScript with full access to app. Treat it as a powerful capability and only run code you trust.
• Stop for exact approval before running untrusted eval, installing/uninstalling plugins, or mutating user content. Use an isolated test vault and a rollback plan.
• Use dev:debug on then attach a CDP client for breakpoint-grade debugging.

See also

• Developer commands reference
• Plugins reference
• Security and authority