Code Humanizer

Run an opinionated AI-slop triage workflow that protects behavior while improving maintainability and review quality.

Workflow

1. Baseline scan.
2. Apply only safe rewrites automatically.
3. Re-run scan and tests.
4. Produce suggest-only recommendations for risky improvements.

Step 1: Baseline Scan

Run:

python skills/code-humanizer/scripts/run_humanizer.py scan <target-path>

Use --json when another tool should consume the result:

python skills/code-humanizer/scripts/run_humanizer.py scan <target-path> --json

Prioritize critical and high issues first, then tackle medium.

Step 2: Safe Rewrites

Preview:

python skills/code-humanizer/scripts/run_humanizer.py rewrite <target-path> --diff

Apply:

python skills/code-humanizer/scripts/run_humanizer.py rewrite <target-path> --apply

Safe rewrites are intentionally narrow:

• Remove low-signal comments that restate obvious behavior.
• Trim trailing whitespace.
• Collapse excessive blank-line runs.

Step 3: Verification

After safe rewrites:

1. Re-run project tests.
2. Re-run scan and compare score trend.
3. Confirm no public behavior contract changed.

Step 4: Suggest-Only Improvements

Do not auto-apply these without explicit approval:

• Rename public identifiers or CLI flags.
• Replace broad exception handling with typed taxonomies.
• Split long functions or large files where structure changes are non-trivial.
• Remove speculative abstractions that may alter extension points.

For each suggest-only item, include:

1. File and line.
2. Why it matters for reviewability/maintainability.
3. Risk level and rollback note.

References

Read only what is needed:

• Core taxonomy and source links: references/ai_slop_taxonomy.md
• Community complaint patterns: references/reddit_patterns.md