Drama Skills
zenstory-ai/drama-skillsEleven skills that take an idea or novel through script, character assets, storyboard, image and video prompts and a rough cut for AI short dramas.
Recommended
Check first Depth 5/5
Claude writing skills change how an agent drafts, edits and formats text. Some set an output style, such as shorter replies. Some give a procedure for a kind of document, such as sales copy or a questionnaire. A few edit existing prose to remove machine-sounding patterns. We reviewed 23: 11 get an install verdict and 12 come with caveats.
Most are short prompts, so judge them by their rules. A 3 KB skill you could write yourself has to earn its place there.
What to look for in Claude writing skills
Look at the rules. Specific instructions beat taste words: “state one claim per paragraph” or “diagnose the input before drafting” can be followed and checked, while “write engagingly” cannot. Then look at whether the skill needs input from you. The better copy skills ask for real product facts first and say what they will not invent.
Next comes scope. Some skills cover a single language or a single format. Others bundle dozens of unrelated skills, and installing everything crowds the list of skill descriptions the agent keeps in context.
On security, writing skills are mostly low risk because they are Markdown. Of the 23, 9 are clean, 10 are marked “review” and 4 have broad access. The “review” label often reflects partial coverage of a large repository, not a specific problem; the security page explains how we assign labels. Watch for installers that write into your agent folders and for skills that call paid APIs.
Recommended writing skills
- copywriting from Corey Haines’ marketing pack is clean and the deepest of the clean picks: a 15.5 KB
SKILL.md, three reference files and ten eval prompts. It diagnoses your input first, then drafts under rules against AI-sounding phrasing. It cannot work without real product facts. - writing-for-agents is clean too. It is a guide to writing skills,
AGENTS.md,CLAUDE.mdand other documents that agents read, covering context pointers and completion criteria. It targets text for agents, not for human readers. - to-questionnaire interviews you about the recipient and what you need back, then writes one Markdown questionnaire. It is clean and small, and our review notes it is simple enough to reproduce yourself.
- caveman is an output-style skill: a 4.7 KB file that makes replies terse and keeps code, paths and errors verbatim. It is Apache-2.0 and has no scripts. It also has about 598,000 recorded installs, so it is a popular way to save tokens. It tells the agent to use plain prose for commit messages and documentation.
- claude-blog is the heavyweight: 65 skills for writing, auditing and refreshing blog posts, with a five gate delivery contract and a pytest suite. An optional installer copies files into your agent skill folders, so read it before running it. Its quality scores are the author’s heuristics, not search-engine data.
Two narrower options deserve a mention. Humanize KR rewrites AI-sounding Korean text with local checker scripts and is useful only if you write Korean. Anti-Defensive Writing is clean but restructures academic papers around the strongest claim, which can hide limitations that reviewers expect to see.
Writing for search overlaps with SEO skills and marketing skills. The category page lists all 23, and how we review covers depth scores.