kmshdev/plugin/smell-detector

Smell Detector

Workflow

1. Confirm the repository root and choose an output directory. Default to

.refactor-code-smells/smell-findings under the target repo when the user does not provide one.

1. Read references/csv-schema.md, references/taxonomy.md,

references/ast-grep-rules.md, and references/refactoring-guru-techniques.md.

1. Resolve the bundled plugin root from the absolute path used to open this SKILL.md,

then run the bundled detector:

``bash skill_file="<absolute-path-to-this-SKILL.md>" plugin_root="$(cd "$(dirname "$skill_file")/../.." && pwd)" python3 "$plugin_root/scripts/detect_smells.py" \ --repo <repo-root> \ --output <repo-root>/.refactor-code-smells/smell-findings ``

Add one --include '<repo-relative-glob>' per requested scan scope. The same include globs constrain structural ast-grep results and text-assisted results.

1. Inspect the printed counts, warnings.log, and a sample from each non-empty CSV before

reporting.

1. Treat every row as a candidate. Verify code context before calling a smell real.
2. Hand the output directory to $smell-refactorer when the user wants fixes.

Output Rules

• Always name the generated CSV directory.
• Report counts per category CSV.
• Report detector warnings; do not treat a failed ast-grep rule as zero findings.
• Distinguish structurally scanned languages from the text-only languages printed by the

detector.

• Mention that candidates came only from source files in the target directory and requested

include scope, and that refactoring still needs verification.

• Do not edit code from this skill unless the user explicitly combines detection and fixing.