| name | prompt-editorial |
| description | Prompt set for Daily Blog scoring, synthesis, highlights, and conservative tone guard. Supports multilingual templates via language-suffixed files. |
Prompt Editorial Skill
Scope
This skill defines runtime prompt templates for Daily Blog:
- Stage-1 scoring (
references/scoring.<lang>.md)
- Stage-2 synthesis (
references/synthesis.<lang>.md)
- Daily highlights (
references/highlights.<lang>.md)
- Style guard rewrite (
references/style-guard.<lang>.md)
- Banned lexicon (
references/banned-lexicon.<lang>.txt)
Runtime behavior
- Application loads prompt files from
skills/prompt-editorial/references.
- Missing or unreadable files fall back to built-in prompts.
- Fallback events are recorded in digest warnings.
Editing rules
- Keep placeholders unchanged:
{{ARTICLES_BLOCK}}
{{SUMMARY}}
{{REASON}}
{{WHY_READ}}
- Keep output contract stable:
- scoring/synthesis/style-guard output strict JSON
- highlights output plain text
- Keep language-specific template content evidence-first and conservative.
- Avoid hype claims, generic marketing copy, and unsupported certainty.
- Recommended language tags:
zh / zh-CN
en
- add more by creating matching
<name>.<lang>.md files.