from claude-essentials109
A practical writing style and persona guide for producing human-sounding docs, READMEs, commit messages, PR descriptions, blog posts, and UI copy.
The Writer skill is a comprehensive style and tone guide designed to help agents produce natural, human-sounding writing across several contexts. It enumerates named personas (Engineer, Architect, PM, Marketer, Educator, Contributor, Outreach Writer, UX Writer), core principles (be concrete, lead with the answer, show reasoning), forbidden patterns (AI tells, em dashes, corporate speak), and formatting advice. The skill is geared toward improving the voice of generated content so it reads like a real person wrote it.
Load this skill when an agent must produce user-facing text: technical documentation, README content, architecture decision records, product specs, blog posts, commit messages, release notes, outreach emails, and short UI copy. Choose the persona that matches the use case (e.g., Engineer for API docs, Marketer for landing pages).
references/ that provide persona-specific examples and guidance (has_references=true)Any text-generation agent or assistant that can load style constraints (Claude, GPT-family, or in-house LLMs) will benefit. This is a content-quality skill rather than a runtime integration. It is most useful when available as a context prompt or preloaded persona for writing tasks.
A writing style guide skill with 8 persona variants (engineer, architect, PM, marketer, educator, contributor, outreach, UX writer) loaded via reference files. No scripts — purely instructional. The SKILL.md is well-structured with clear triggers, forbidden patterns, and formatting rules. Minor security deduction for referencing external persona files without integrity checks, but no executable code risk. Well-written and practical for anyone producing human-sounding content.
Clean skill with no executable code. The persona selection table is a nice pattern for multi-mode skills. Forbidden patterns section is well-curated and specific. Missing the actual persona reference files in the fetched data, so completeness can't be fully assessed.
