Codex Iterative Fix Skill

LIBRARY-FIRST PROTOCOL (MANDATORY)

Step 1: Library Catalog

• Location:

  .claude/library/catalog.json

• If match >70%: REUSE or ADAPT

Step 2: Patterns Guide

• Location:

  .claude/docs/inventories/LIBRARY-PATTERNS-GUIDE.md

• If pattern exists: FOLLOW documented approach

Step 3: Existing Projects

• Location:

  D:\Projects\*

• If found: EXTRACT and adapt

Decision Matrix

Kanitsal Cerceve (Evidential Frame Activation)

Purpose

When to Use This Skill

• Multiple tests failing that need iterative fixing
• Debugging issues requiring trial-and-error
• Refactoring with test validation
• CI/CD failures needing automated fixes
• Build errors requiring multiple attempts

When NOT to Use This Skill

• Research tasks (use multi-model-discovery)
• Understanding codebase (use gemini-codebase-onboard)
• Critical production code (use sandbox mode first)
• Architecture decisions (use llm-council)

Workflow

Phase 1: Initial Assessment

1. Identify failing tests/errors
2. Determine scope of fixes needed
3. Choose appropriate mode:

   • full-auto

     : Standard autonomous mode

   • sandbox

     : For risky changes

   • yolo

     : When speed is critical

Phase 2: Codex Execution

# Standard iterative fix
./scripts/multi-model/codex-yolo.sh "Fix all failing tests" task-id "." 15 full-auto

# With sandbox protection
./scripts/multi-model/codex-yolo.sh "Fix tests" task-id "." 10 sandbox

# Via delegate.sh
./scripts/multi-model/delegate.sh codex "Fix all failing tests and verify they pass" --full-auto

Phase 3: Verification

1. Codex runs tests after each fix
2. Iterates until all tests pass
3. Claude reviews final changes
4. Summary of what was fixed

Success Criteria

• All targeted tests passing
• No new regressions introduced
• Changes reviewed and validated
• Clear documentation of fixes

Example Usage

Example 1: Test Suite Failures

User: "CI is failing with 12 test errors"

Codex Process:
1. Run tests, capture failures
2. Analyze first failure
3. Implement fix
4. Re-run tests
5. Repeat until all pass

Output:
- Fixed: 12 tests
- Changes: 8 files modified
- Root causes: Missing null checks, outdated mocks

Example 2: Type Errors

User: "TypeScript build has 47 type errors"

Codex Process:
1. Run tsc, capture errors
2. Fix type errors systematically
3. Re-run after each batch
4. Verify build succeeds

Output:
- Fixed: 47 type errors
- Patterns: Missing types, incorrect generics
- Added: 3 new type definitions

Modes Explained

--full-auto

--sandbox workspace-write

--yolo

Integration with Meta-Loop

META-LOOP IMPLEMENT PHASE:
    |
    +---> codex-iterative-fix
    |         |
    |         +---> Codex: Run tests
    |         +---> Codex: Fix failures
    |         +---> Codex: Iterate until pass
    |
    +---> Continue to TEST phase (verification)

Memory Integration

• Key:

  multi-model/codex/iterative-fix/{project}/{task_id}

• Tags: WHO=codex-iterative-fix, WHY=autonomous-fixing

Invocation Pattern

# Via router (automatic detection)
./scripts/multi-model/multi-model-router.sh "Fix all failing tests"

# Direct Codex call
bash -lc "codex --full-auto exec 'Fix all failing tests and verify they pass'"

Guardrails

• Run on production without review
• Skip final verification
• Ignore new test failures
• Exceed iteration limit without human check

• Review changes before commit
• Document what was fixed
• Check for regressions
• Store results in Memory-MCP

Related Skills

• codex-safe-experiment

  : Sandbox experimentation

• smart-bug-fix

  : Systematic debugging

• testing-quality

  : Test generation

• multi-model-discovery

  : Find existing fixes

Verification Checklist

• All targeted tests passing
• No new regressions
• Changes reviewed
• Memory-MCP updated
• Documentation complete