Reference

`peekaboo learn`

peekaboo learn

peekaboo learn prints the canonical “agent guide” that powers Peekaboo’s AI flows. It stitches together the generated system prompt, every tool definition from ToolRegistry, best-practice checklists, common workflows, and the full Commander signature table so other runtimes can stay in sync with the CLI release.

#What it emits

  • System instructions straight from AgentSystemPrompt.generate(), including communication rules and safety guidance.
  • Tool catalog grouped by category with each tool’s abstract, required/optional parameters, and JSON examples (if available).
  • Best practices + quick reference: long-form guidance for automation patterns, then a condensed cheat sheet.
  • Commander section: a programmatic dump of every CLI command’s positional arguments, options, and flags (built by CommanderRegistryBuilder.buildCommandSummaries()).

#Implementation notes

  • The command is intentionally text-only—--json is ignored—so downstream systems should capture stdout if they want to cache the content.
  • Everything runs on the main actor because it pulls live data from ToolRegistry and Commander; no stale handwritten docs are involved.
  • Because it reuses the same builders the CLI uses at runtime, new commands/tools automatically show up here as soon as they land.
  • When stdout is a rich TTY, output is rendered with Swiftdansi for ANSI color and table/box formatting; piped output stays plain Markdown for downstream tools.

#Examples

# Save the full guide for another agent runtime
peekaboo learn > /tmp/peekaboo-guide.md

# Extract just the Commander signatures
peekaboo learn | awk '/^## Commander/,0'

#Troubleshooting

  • Verify Screen Recording + Accessibility permissions (peekaboo permissions status).
  • Confirm your target (app/window/selector) with peekaboo list/peekaboo see before rerunning.
  • Re-run with --json or --verbose to surface detailed errors.