Skillsmith
Expert architect for designing production-quality Agent Skills. Use when creating new skills, refactoring existing skills, converting reusable prompts into skills, or designing modular skill collections.From its SKILL.md
npx -y skills add ViGi-P/skillsmith --skill skillsmithAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
3.6 KB, 659 tokens by cl100k_base, as published. Nobody here has run it
Skillsmith
Design Agent Skills as reusable software components—not as large prompts.
Your objective is to produce skills that are discoverable, composable, maintainable, and token-efficient.
If official documentation is available, treat it as the source of truth. Otherwise, follow the provided specification.
Principles
| Prefer | Avoid |
|---|---|
| Single responsibility | God skills |
| Progressive disclosure | Bloated SKILL.md files |
| Focused references | Duplicate documentation |
| Modular design | Overlapping responsibilities |
| Searchable descriptions | Generic descriptions |
| Scripts for executable logic | Large code blocks in markdown |
| Assets for reusable resources | Embedded templates or schemas |
| Deterministic workflows | Ambiguous instructions |
| Token efficiency | Repetition |
Workflow
1. Evaluate
Determine whether the request should become a skill.
A skill should represent a reusable capability.
If the request is better solved by a script, template, documentation, or a one-time prompt, recommend that instead.
Define:
- responsibility
- intended users
- activation criteria
- inputs
- outputs
- constraints
- dependencies
- edge cases
If multiple unrelated capabilities exist, split them into separate skills.
2. Design
Design the directory structure before writing files.
Create only the directories that add value.
Typical layout:
skill-name/
├── SKILL.md
├── references/
├── scripts/
└── assets/
Decide where each piece of information belongs.
SKILL.md
Keep concise.
Include only:
- activation criteria
- workflow
- important rules
- validation
- references to supporting files
references/
Store:
- technical documentation
- APIs
- algorithms
- schemas
- FAQs
- workflows
scripts/
Store executable logic.
Scripts should validate inputs, document dependencies, and fail predictably.
assets/
Store reusable resources such as:
- templates
- sample data
- configuration files
- diagrams
- example documents
3. Implement
Generate:
- directory tree
- valid frontmatter
- concise
SKILL.md - supporting reference files
- scripts when appropriate
- assets when appropriate
Descriptions should explain:
- what the skill does
- when another agent should activate it
Assume the description will be used for discovery.
4. Review
Before finalizing, verify:
- specification compliant
- single responsibility
- discoverable
- modular
- maintainable
- progressively disclosed
- minimal duplication
- token efficient
Validate against:
references/constraints.md
If any issue is found, revise before returning the skill.
Output
Unless instructed otherwise, always provide:
- Complete directory tree
- Complete contents of every generated file
- Valid relative file references
- Brief design notes explaining important architectural decisions
Final Check
Before returning the skill, ask:
If this were published in a large public registry, would another agent reliably discover it, understand when to activate it, compose it with other skills, and successfully execute it without additional explanation?
If the answer is not an unqualified yes, continue refining the design.