/codex "implement the UserProfile component per the spec"
/codex "write unit tests for the auth module"
/codex "refactor database queries to use parameterized statements"
/codex --dry-run "rename all snake_case vars to camelCase in src/utils"

RUN: command -v codex
IF: not found
  OUTPUT: "Codex CLI not installed. Install with: npm install -g @openai/codex"
  END

RUN: codex --version
OUTPUT: "Codex CLI v{version} detected"

IF: --dry-run flag present
  DRY_RUN: true
  TASK: remaining arguments after --dry-run

IF: no arguments provided
  Analyze current context (git diff, recent files, branch name)
  Suggest what could be delegated to Codex
  OUTPUT: "No task specified. Based on current context, consider delegating:"
  OUTPUT: "  - {suggestion based on uncommitted changes}"
  OUTPUT: "  - {suggestion based on branch name}"
  END

GATHER context:
  - Working directory and project structure
  - Relevant source files (read files mentioned in task or inferred from context)
  - Project coding patterns (from existing code)
  - Test patterns (from existing tests)

CONSTRUCT prompt:
  - Task description from user args
  - File list to modify
  - Coding standards observed in project
  - Any constraints or patterns to follow

IF: DRY_RUN
  OUTPUT: "Dry Run - Would execute:"
  OUTPUT: "  Prompt: {constructed prompt}"
  OUTPUT: "  Files in scope: {file list}"
  OUTPUT: "  Estimated changes: {description}"
  END

RUN: codex exec "$PROMPT" --json
CAPTURE: output as JSON

PARSE: JSON output from Codex
EXTRACT: files changed, diffs applied

OUTPUT:
  "Codex Execution Complete"
  ""
  "Files changed:"
  "  - {file1}: {description of change}"
  "  - {file2}: {description of change}"
  ""
  "Diff summary:"
  {abbreviated diff output}

DETECT: test runner (npm test, pytest, go test, etc.)

IF: test runner found
  RUN: tests
  IF: tests pass
    OUTPUT: "All tests passing after Codex changes"
  ELSE:
    OUTPUT: "Test failures detected — review Codex output"
    OUTPUT: {failure details}

IF: no test runner
  OUTPUT: "No test runner detected — manual verification recommended"

OUTPUT:
  "/codex Complete"
  ""
  "| Metric | Value |"
  "|--------|-------|"
  "| Files changed | {count} |"
  "| Lines added | {added} |"
  "| Lines removed | {removed} |"
  "| Tests | {pass/fail/none} |"

User: /codex "add input validation to the signup form"

Codex CLI v0.1.0 detected
Building execution context...

Reading: src/components/SignupForm.tsx
Reading: src/utils/validation.ts
Reading: tests/signup.test.ts

Executing via Codex CLI...

Codex Execution Complete

Files changed:
  - src/components/SignupForm.tsx: Added validation for email, password, name fields
  - src/utils/validation.ts: Added validateSignupInput function

Diff summary:
  src/components/SignupForm.tsx | 15 ++++++++++++---
  src/utils/validation.ts      |  8 ++++++++

Running tests...
All tests passing after Codex changes

/codex Complete

| Metric | Value |
|--------|-------|
| Files changed | 2 |
| Lines added | 23 |
| Lines removed | 3 |
| Tests | pass |

User: /codex --dry-run "rename all snake_case vars to camelCase in src/utils"

Codex CLI v0.1.0 detected

Dry Run - Would execute:
  Prompt: "Rename all snake_case variables to camelCase in the following files..."
  Files in scope: src/utils/helpers.ts, src/utils/format.ts, src/utils/api.ts
  Estimated changes: Variable renames across 3 files

Ready to execute? Run without --dry-run