Candid Loop

Run code review repeatedly until all issues are resolved.

Usage

/candid-loop

Candid Loop automates the fix-review-fix cycle. It runs candid-review, applies fixes, and repeats until your code is clean (or max iterations is reached).

Modes

Auto Mode (Default)

/candid-loop

Automatically applies all fixes without prompting. Best for:

Fast iteration during development
CI/CD pipelines
Pre-commit hooks

Review-Each Mode

/candid-loop --mode review-each

Go through each fix one by one with simple Yes/No prompts. Best for:

Understanding what issues exist
Learning the codebase
Quick selective fixing

Interactive Mode

/candid-loop --mode interactive

Full control with skip, ignore, and batch options. Best for:

Building an ignore list for false positives
Complex review sessions
When you need to skip entire batches

How It Works

[1/5] Run candid-review → Found 3 issues → Apply fixes
[2/5] Run candid-review → Found 1 issue → Apply fix
[3/5] Run candid-review → No issues found → Done!

The loop continues until:

No issues remain (success)
Max iterations reached (stops with warning)
User cancels (interactive mode only)

Configuration

CLI Flags

Flag	Description	Default
`--mode <auto\|review-each\|interactive>`	Execution mode	`auto`
`--max-iterations <N>`	Maximum loop iterations	`5`
`--categories <list>`	Categories to enforce	`all`

Config File

Add to .candid/config.json:

{
  "loop": {
    "mode": "auto",
    "maxIterations": 5,
    "enforceCategories": ["all"],
    "ignored": {
      "categories": [],
      "patterns": [],
      "ids": []
    }
  }
}

Ignoring Issues

Prevent specific issues from blocking your loop.

Ignore by Category

Skip entire categories of issues:

{
  "loop": {
    "ignored": {
      "categories": ["edge_case", "architectural"]
    }
  }
}

Ignore by Pattern

Skip issues matching title patterns (regex):

{
  "loop": {
    "ignored": {
      "patterns": ["Unicode", "timezone", "DST"]
    }
  }
}

Ignore by ID

Skip specific issues by their unique ID (found in .candid/last-review.json):

{
  "loop": {
    "ignored": {
      "ids": ["a1b2c3d4e5f6"]
    }
  }
}

Tip: In interactive mode, you can choose “Add to ignore list” for any issue to automatically add its ID to the config.

Enforcing Specific Categories

Only fix certain types of issues:

/candid-loop --categories critical
/candid-loop --categories critical,major

Valid categories: critical, major, standards, smell, edge_case, architectural, all

Examples

# Default: auto-fix all issues, max 5 iterations
/candid-loop
 
# Review-each: go through fixes one by one
/candid-loop --mode review-each
 
# Interactive: full control with skip/ignore options
/candid-loop --mode interactive
 
# Quick pass: only fix critical issues, max 3 tries
/candid-loop --categories critical --max-iterations 3
 
# Thorough: fix critical and major, up to 10 iterations
/candid-loop --categories critical,major --max-iterations 10

Output Examples

Success

[Auto Mode] Running candid-loop with max 5 iterations...

[1/5] Found 3 issues. Applying fixes...
      ✓ Fixed: Null check missing in user.ts:42
      ✓ Fixed: SQL injection in db.ts:15
      ✓ Fixed: N+1 query in orders.ts:88

[2/5] Found 1 issue. Applying fix...
      ✓ Fixed: Missing error handling in auth.ts:20

[3/5] No issues found!

Summary:
- Iterations: 3
- Issues fixed: 4
- Status: PASS

Max Iterations Reached

[Auto Mode] Running candid-loop with max 5 iterations...

[5/5] Found 2 issues. Applying fixes...
      ✓ Fixed: Issue A
      ✓ Fixed: Issue B

Max iterations (5) reached. 1 issue remains.

Remaining issues:
  💭 Complex architectural concern in core.ts:100

Consider:
- Increasing --max-iterations
- Adding to ignore list for false positives
- Manually reviewing complex issues

Combining with Other Features

Candid Loop respects your other settings:

# Use harsh tone in the underlying reviews
/candid-loop  # (with "tone": "harsh" in config)

# Focus reviews on security issues only
/candid-loop --categories critical