How to use
- Copy the installation command and review the source package before installing it in your own agent environment.
- Check the source document for required tools, credentials and supported hosts.
- Describe your input and desired deliverable; review the output against the source material.
npx skills add https://github.com/sylphxai/skills --skill author-skillBefore you start
- Installation does not include a model subscription, API allowance or permission to access your files.
- Repository stars and installation counts describe activity, not an independent quality or security test.
Rendered from the source repository. Headings, examples, code, tables, links, and referenced images are preserved.
Author Skill
Skills exist because model knowledge lags. Training still produces 2024-style MUST/NEVER recipes, numbered ordinary work, and subagent prompts that copy the parent system prompt. Current capable models need particular gotchas and judgement, not a second textbook. A longer body is not a better skill.
Prefer one heuristic that stays true: match surrounding code, naming, and idiom. Constrain only where a wrong call is expensive — money, deletion, credentials, safety, or a public contract.
A skill description is the trigger: what it does, when to use it, phrases a user would type, and nearby cases that should not fire. All "when" belongs there, not in the body.
The body encodes opinions or facts the model would otherwise miss. Do not write a numbered recipe of ordinary work, Complete-when checklists, or output noun piles. Do not add verification steps a capable model already runs. A subagent prompt is the same job at a narrower scope — gotchas for that job, not a dump of the parent system prompt. Prefer a script or a short template when the job is fragile or needs a format. Put long material in references/ and say when to open each file. Say each fact once.
For a skill package: the folder name matches name; every local link resolves. A public skill carries reusable method only — not owner-only procedure, secrets, or private topology. Restore a deleted job only when the requestable artifact and a unique mechanism still exist.
Open context engineering when writing a system prompt, CLAUDE.md, AGENTS.md, or subagent prompt, or when deciding what belongs in a skill versus those layers.
Use curate-skill-repository for portfolio-wide overlap, recovery, or retirement.
