Reusable skills for Claude (Cowork, Claude Code, and compatible agents). One folder per skill; each folder contains a SKILL.md and optional references/, scripts/, and assets/.
| Skill | Language | What it does |
|---|---|---|
slop-checker |
English | Scores outbound documents (.docx/.pdf/.pptx/drafts) for AI-generated slop. Two modes: CT (Twitter) and Professional. Read-only by default — reports and recommends, never edits without permission. |
chinese-slop-checker |
中文 | 检测中文文本的AI味，按平台（微信公众号/小红书/抖音/通用）评分并给出修改建议。默认只读。 |
With the Skills CLI:
npx skills add https://github.com/viewfin-labs/skills --skill slop-checker
npx skills add https://github.com/viewfin-labs/skills --skill chinese-slop-checkerIn Cowork: Settings → Capabilities → add skill from folder.
Some skills reference internal data sources. Placeholders you may need to fill in after install:
<YOUR_NOTION_KNOWLEDGE_PAGE_ID>— both slop checkers fetch their latest rule framework from a Notion Knowledge page. Without it, they fall back to the bundledreferences/copy.
- One folder per skill, kebab-case, named exactly as the skill's
name:frontmatter. SKILL.mdis mandatory — frontmattername+description(the description is the trigger surface; write it for the model, with concrete trigger phrases).references/— wordlists, frameworks, examples the skill reads at runtime. Bundle a fallback copy of anything normally fetched from an external source, and date it.scripts/— deterministic helpers (scanners, converters). Prefer scripts over LLM judgment for anything countable.- No secrets or internal IDs in committed files. Use self-documenting placeholders (
<YOUR_..._ID>) and document them in the Configuration section above. - Versioning: note the version and date in the skill body or reference files; tag repo releases (
v2026.07) when batching updates.
- Create
<skill-name>/SKILL.md(+references/,scripts/as needed). - Sweep for secrets and internal identifiers before committing.
- Add a row to the Skills table above.