Todosweep: every TODO found, ranked and triaged

Codebases inevitably accumulate TODO, FIXME, and HACK comments that eventually turn into invisible technical debt. This skill scans your repository to extract, categorize, and triage these markers, providing a clear map of what needs immediate fixing versus what can be safely ignored or deleted.

What it does

• Multi-tag extraction scans for TODO, FIXME, HACK, BUG, XXX, NOTE, and WARNING tags across your files.
• Urgency triaging automatically ranks findings into High, Medium, Low, and Info categories based on tag type.
• Git integration uses git blame to show who last edited each line and how long ago.
• Stale detection flags lines untouched for more than 30 days as stale to help identify forgotten or irrelevant tasks.
• Contextual filtering allows narrowing scans by subfolders, specific files, authors, or tag types.

How it works

1. Targeting: Specify a directory or file using Todosweep <path>.
2. Filtering: Refine the results by adding criteria like "only stale ones" or "by author-name".
3. Analysis: The skill runs a scan while ignoring node_modules and build directories.
4. Triage: Review the prioritized list and specific suggestions for addressing, converting, or deleting items.

Frameworks & tools

This tool relies on git blame for metadata and works across any language that uses standard comment syntax, including TypeScript, JavaScript, Python, and Go.

Why this beats prompting it yourself

Generic prompts often miss the specific urgency hierarchy and fail to integrate git blame data for age-based triaging. This skill enforces a structured output that separates high-risk bugs from low-priority notes while automatically filtering out noise from vendor directories.

Use cases

• Release preparation to ensure no BUG or FIXME tags remain in the production path.
• Tech debt audits to identify HACK comments that have been stale for months.
• Onboarding to quickly map out the known issues and architectural warnings in a new codebase.

Known limitations

Scans only code comments and skips documentation files by default. Age calculation is based on the last time the specific line was edited, not necessarily when the comment was first created.