Skip to content

xlflow inspect-gui ​

Scan source for automation-hostile GUI boundaries without opening Excel.

Usage ​

bash
xlflow inspect-gui

Options and Arguments ​

Option / argumentDescriptionDefault
--jsonReturn detected boundaries and source locations.false

Examples ​

bash
xlflow inspect-gui
xlflow inspect-gui --json

Notes ​

TIP

Run inspect-gui before headless automation to find likely MsgBox, InputBox, dialog, or UserForm boundaries.

WARNING

This is static analysis. It is intentionally conservative and does not prove a macro is fully headless-safe.

JSON Output Example ​

Successful --json output uses the xlflow envelope plus command-specific fields.

json
{
  "status": "ok",
  "command": "inspect-gui",
  "boundaries": [{ "file": "src/modules/Main.bas", "line": 12, "kind": "MsgBox" }]
}

When to use this command ​

Use xlflow inspect gui when the task matches the command description above. For a goal-oriented workflow, start with the How-to guides and return here for exact options.

Prerequisites ​

Check the project configuration and run xlflow doctor --json before workbook-backed operations. Source-only commands can run without Excel; commands that read or mutate a workbook require Windows Excel and VBIDE access.

What this command reads and changes ​

The command reads the inputs and configuration described in its syntax and examples. Treat source files, the saved workbook, and a live session as separate states; add --session when the live workbook is authoritative. Any mutation is reversible only when a backup or explicit session save boundary exists.

Effect on source-of-truth state ​

Use xlflow status --json before and after the command. A source edit normally requires push; a workbook edit normally requires pull; a dirty live session requires save --session or an intentional discard.

Common workflows ​

Combine this command with the relevant source/workbook/session workflow, and use --json in scripts and agent loops.

Common failures ​

Read the structured error.code, exit code, and recovery metadata instead of scraping terminal text. The symptom-oriented troubleshooting guide maps installation, execution, session, VS Code, and WSL failures to recovery steps.

Released under the MIT License.