sniff

name	sniff
description	Runs structured code-smell analysis and maps findings to standard smell categories. Use when prioritizing maintainability risks and refactors.
user-invocable	true
disable-model-invocation	false

Sniff

Core Contract

Run evidence-based code-smell review and map findings to reference smell IDs. Use ./smells/index.json plus ./smells/<category>/<smell-id>.md. Follow CLAUDE.md / AGENTS.md on conflict.

Required Inputs

1. Review scope.
2. Language/framework context.
3. Strictness level.
4. Reference-only mode (default true).

Workflow

1. Frame review goal and choose 2-4 priority smell categories.
2. Load the reference smell hierarchy and resolve IDs.
3. Produce evidence-backed findings with related refactor suggestions.
4. Rank severity (high/medium/low) and include one intentional non-finding.
5. Verify every finding maps to a reference ID and concrete code evidence.

Stop and ask before introducing custom smell labels or proposing out-of-scope architecture changes.

Safety Rules

• Never label something a smell without evidence in the reviewed code.
• Never demand broad refactors when a localized fix addresses the risk.
• Never force category coverage; report only smells that appear.
• Never invent smell IDs outside ./smells/index.json without user approval.

Output Style

Report scope, evaluated categories/smells, grouped findings with evidence and related refactors, severity-ranked top actions, and one intentional non-finding.