Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.ironbee.ai/llms.txt

Use this file to discover all available pages before exploring further.

When you run ironbee install, IronBee installs a set of slash commands that the AI agent can invoke directly inside Claude Code or Cursor. These commands guide the agent through verification and analysis workflows.

​
/ironbee-verify

Run a verification cycle for the current changes. 
/ironbee-verify
Default behavior: focused verification — the agent reviews what changed and tests only the affected areas. This is the most common mode and the one IronBee triggers automatically at task completion. The agent will:
  1. Identify which files changed and what cycles are active for this project
  2. Start verification and exercise every active cycle (browser, Node, or backend)
  3. Navigate pages, run probes, or make API calls as required
  4. Check for console errors, network failures, and unexpected responses
  5. Submit a single verdict with evidence for all active cycles

​
/ironbee-verify full

Complete verification — runs every checklist item regardless of what changed. 
/ironbee-verify full
Use this when:
  • You want comprehensive coverage beyond just the changed areas
  • You are doing a final check before a release
  • A change is broad enough that focused verification might miss regressions

​
/ironbee-verify visual

Visual-only verification — checks UI appearance without testing functionality. 
/ironbee-verify visual
Covers:
  • Contrast and accessibility
  • Layout and spacing
  • Typography and fonts
  • Images and icons
  • Theming (light/dark mode)
Use when changes are purely cosmetic — CSS, layout, design tokens.

​
/ironbee-verify functional

Functional-only verification — tests behavior without visual checks. 
/ironbee-verify functional
Covers:
  • Click interactions and event handlers
  • Form submission and validation
  • Navigation and routing
  • Data flow and state updates
  • Error handling and edge cases
Use when changes are logic or behavior changes with no visual impact.

​
/ironbee-analyze

Run session analytics and get LLM-powered semantic insights for the current session. 
/ironbee-analyze
The agent will:
  1. Pull session metrics (time spent, tool calls, pass/fail counts)
  2. Identify patterns — which files caused failures, where retries clustered
  3. Surface concrete recommendations
Output is printed inline in the agent conversation. The same analysis is also available in the Console.

​
When are these triggered automatically?

IronBee installs a Stop hook that fires when the agent tries to complete a task. If code files were changed, the hook automatically triggers the verification flow (equivalent to /ironbee-verify) before allowing the task to finish. Every active cycle — browser, Node, backend — must pass within the same verification attempt before the task can complete. You can also invoke any of these commands manually at any time — for example, to run a quick visual check mid-task or get early analysis.