Docs

AI Commands

Agentastic supports slash commands in launcher input. Some commands are parsed structurally (with dedicated extractors), and some are treated as mode hints.

Structured Tags

These have explicit parsing logic:

  • /model:<value>
  • /screen:<active|0|1..n>
  • /search (optional inline query)
  • /imagine
  • /tab: (opens tab selector)
  • /tool:<group-id>
  • /skill:<skill-name>
  • /app:<app-name>
  • /app:<app-name>:<subcommand>

Selectors open when input ends with one of:

  • /tab:
  • /model:
  • /tool:
  • /app:
  • /skill:

Mode Hints

These are recognized in prompt text to influence mode:

  • /agent or /agentastic -> enables agent mode
  • /think or /thinking -> requests thinking mode
  • /search -> enables search mode for the turn

Basic Examples

/model:openai:gpt-5 Summarize this issue briefly
/screen:2 Explain this chart
/search macOS accessibility api changes
/tab: Compare these docs
/tool:email Draft a customer follow-up
/skill:security-review Audit this plan
/app:slide Build a 5-slide executive summary
/app:slide:open

How Tags Affect Prompt Text

  • Parsed tags are removed from the message body before request assembly.
  • /tab: inserts a visible placeholder in your input and adds a browser context block at send time.
  • /tool: updates explicit tool selection state for the conversation.

Next