Claude Skills by vkroz

Add explanatory comments to code to improve readability and maintainability. Use when the user asks to "comment code", "add comments", "document this code", "explain this code with comments", or wants inline documentation added to source files.

This skill should be used when the user asks to "commit", "commit changes", "create a commit", "stage and commit", "git commit", or wants to save their work to version control. Guides Claude through staging files and writing a well-formed git commit message.

Review a requested code scope for concrete defects and maintainability problems; report evidence without modifying implementation.

Create requirements and scoped work items in Taskman, from product direction to an epic, feature or task.

Refine the design of an existing product, epic or feature while preserving its agreed outcome.

Review project documentation for drift against completed Taskman work and update affected documents.

Execute authorized Taskman work through implementation, local verification and recorded completion.

Decompose an existing Taskman epic or feature into executable work with dependencies.

Check work against acceptance criteria and record actual evidence in Taskman. Use for task, feature, epic or product validation.

Users need a skill that processes the doc-sync queue, reviews upstream docs against the closed unit's body and linked spec, proposes diffs, and waits for user approval before applying changes. The skill must follow `rules/doc-maintenance.md` for what to check on closure.

`/agn:design epic` and `/agn:design feature` are placeholders pending the Planner. They should perform focused design revisions on existing units by delegating to the Planner in refine mode.

`/agn:plan epic` and `/agn:plan feature` are placeholders pending the Planner. They should re-decompose an existing epic into features or an existing feature into tasks by delegating to the Planner in refine mode with plan-only intent.

`/agn:validate task` is a placeholder. It should be a light, main-session check that runs the task file's `## Quality gates` end-to-end, reports results, and (optionally) offers to close the task. No QA sub-agent — gates are short and the parent session has the right context.