Use when starting genuinely new work: a new feature, a new component/module, a new script, "build me a...", "add support for...", "create a...", or any request with no existing partial implementation to iterate on. Do not use for bug fixes, small edits, follow-ups, or requests continuing work already in progress.
Installs into .claude/skills of the current project.
Are you the author of Fresh Work?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/joe-signorile-fresh-work)
---
name: fresh-work
description: >
Use when starting genuinely new work: a new feature, a new component/module,
a new script, "build me a...", "add support for...", "create a...", or any
request with no existing partial implementation to iterate on. Do not use
for bug fixes, small edits, follow-ups, or requests continuing work already
in progress.
---
Before writing any code:
1. Confirm scope in a compact key:value block, not prose:
```
goal: <what>
constraints: <what must hold>
files: <relevant existing files, if any>
delegation: <tier per piece of work, or "self">
```
2. Ask clarifying questions if genuinely ambiguous — don't silently guess on
anything that would materially change the design.
3. If Claude Code's plan mode is already active, defer to it — this skill
covers fresh work requested outside an explicit plan-mode session.
4. Once scope is confirmed, execute without restating the plan or
narrating each step (claudia's anti-ceremony rule still applies).
This supplements, not replaces, the claudia output style's voice,
minimalism ladder, and delegation ladder — keep using all three.
