Skip to content

case-library.md

Latest commit

 

History

History
222 lines (148 loc) · 6.11 KB

case-library.md

File metadata and controls

222 lines (148 loc) · 6.11 KB
title Case Library
description Official reproducible scenarios that show what RexCLI can do in real workflows.

Official Case Library

This page is the canonical capability map for RexCLI.

Each case includes:

  • When to use: decision trigger
  • Run: copy-paste commands
  • Evidence: what proves success

Featured English Deep Dives

Star on GitHub{ .md-button .md-button--primary data-rex-track="cta_click" data-rex-location="case_library_featured" data-rex-target="github_star" } Raw CLI vs RexCLI Layer{ data-rex-track="cta_click" data-rex-location="case_library_featured" data-rex-target="compare_workflows" } Case: Cross-CLI Handoff{ data-rex-track="cta_click" data-rex-location="case_library_featured" data-rex-target="case_handoff" } Case: Browser Auth-Wall Flow{ data-rex-track="cta_click" data-rex-location="case_library_featured" data-rex-target="case_authwall" } Case: Privacy Guard Config Read{ data-rex-track="cta_click" data-rex-location="case_library_featured" data-rex-target="case_privacy" }

Case 1: 5-minute fresh setup on a new machine

When to use

You are onboarding a new laptop or teammate and need a clean baseline quickly.

Run

scripts/setup-all.sh --components all --mode opt-in
scripts/verify-aios.sh
powershell -ExecutionPolicy Bypass -File .\scripts\setup-all.ps1 -Components all -Mode opt-in
powershell -ExecutionPolicy Bypass -File .\scripts\verify-aios.ps1

Evidence

  • verify-aios exits with code 0
  • doctor-* checks show no blocking errors

Case 2: Browser MCP installation and smoke test

When to use

You need browser automation (browser_*) working for demos or agent workflows.

Run

scripts/install-browser-mcp.sh
scripts/doctor-browser-mcp.sh

Then in client chat:

browser_launch {"profile":"default"}
browser_navigate {"url":"https://example.com"}
browser_snapshot {"includeAx":true}
browser_close {}

Evidence

  • doctor-browser-mcp reports Result: OK (warnings are acceptable)
  • Smoke commands return structured tool responses without runtime exceptions

Case 3: Cross-CLI handoff in one project

When to use

You want Claude to analyze, Codex to implement, and Gemini to review without losing context.

Run

claude
codex
gemini

Or deterministic one-shot:

scripts/ctx-agent.sh --agent claude-code --prompt "Summarize blockers and propose next steps"
scripts/ctx-agent.sh --agent codex-cli --prompt "Implement the top priority fix from latest checkpoint"
scripts/ctx-agent.sh --agent gemini-cli --prompt "Review risk and missing tests"

Evidence

  • New session/checkpoint artifacts under memory/context-db/
  • Later CLI runs can continue using the same project context

Case 4: Auth-wall handling (human-in-the-loop)

When to use

Automation reaches login walls (Google, Meta, platform auth) and should not blindly bypass them.

Run

browser_launch {"profile":"local"}
browser_navigate {"url":"https://target.site"}
browser_auth_check {}

If requiresHumanAction=true, complete login manually in that browser profile, then continue with browser_snapshot / browser_click / browser_type.

Evidence

  • browser_auth_check returns explicit auth state fields
  • Flow resumes after manual login using the same profile

Case 5: One-shot auditable execution chain

When to use

You need one command to produce an auditable record (init -> session -> event -> checkpoint -> pack).

Run

scripts/ctx-agent.sh --agent codex-cli --project RexCLI --prompt "Continue from latest checkpoint and execute next step"

Evidence

  • New checkpoint entry in memory/context-db/index/checkpoints.jsonl
  • Exported context packet in memory/context-db/exports/

Case 6: Team skill lifecycle (install/update/doctor/uninstall)

When to use

You manage shared skills across multiple CLIs and need predictable lifecycle operations.

Run

scripts/install-contextdb-skills.sh
scripts/doctor-contextdb-skills.sh
scripts/update-contextdb-skills.sh
# rollback if needed
scripts/uninstall-contextdb-skills.sh
powershell -ExecutionPolicy Bypass -File .\scripts\install-contextdb-skills.ps1
powershell -ExecutionPolicy Bypass -File .\scripts\doctor-contextdb-skills.ps1
powershell -ExecutionPolicy Bypass -File .\scripts\update-contextdb-skills.ps1
# rollback if needed
powershell -ExecutionPolicy Bypass -File .\scripts\uninstall-contextdb-skills.ps1

Evidence

  • Doctor output confirms expected targets exist and are healthy
  • Update/uninstall produce no dangling broken links

Case 7: Shell wrapper recovery and rollback

When to use

A user reports command wrapping issues and you need a safe recover path.

Run

scripts/doctor-contextdb-shell.sh
scripts/update-contextdb-shell.sh
# full rollback if needed
scripts/uninstall-contextdb-shell.sh
powershell -ExecutionPolicy Bypass -File .\scripts\doctor-contextdb-shell.ps1
powershell -ExecutionPolicy Bypass -File .\scripts\update-contextdb-shell.ps1
# full rollback if needed
powershell -ExecutionPolicy Bypass -File .\scripts\uninstall-contextdb-shell.ps1

Evidence

  • Wrapper doctor no longer reports blocking issues
  • Native codex/claude/gemini commands work after rollback

Case 8: Security hygiene pre-release check

When to use

Before publishing updates, verify no unsafe config drift in skills/hooks/MCP settings.

Run

scripts/doctor-security-config.sh
powershell -ExecutionPolicy Bypass -File .\scripts\doctor-security-config.ps1

Evidence

  • Security doctor exits 0
  • Any warnings are reviewed and resolved before release

Contribute a new official case

To propose a case for this library:

  1. Include exact commands with no placeholders.
  2. Define measurable evidence (exit code, file artifact, or tool response).
  3. Add rollback/recovery step when relevant.