Content creation skill

Media Content Creation Package

Attribution

Created and maintained by Chuluu.

• Copyright: Copyright (c) 2026 Chuluu
• License: MIT
• GitHub publisher: ChuluuMGL
• Website: https://github.com/ChuluuMGL

Turn one source material into a platform-ready media content package: purpose, drafts, score, targeted revisions, platform formats, tags, cover-image prompts, and a manual publishing checklist.

Do not position this skill as a legal, academic, medical, or formal research-writing skill. For those domains, keep the output as media commentary or ask for a domain-specific workflow.

Scope

Included	Not included by default
Media posts, creator essays, brand articles, tutorials, case recaps, SEO articles, short-video scripts	Legal documents, academic papers, regulatory advice, automatic platform publishing
Purpose-led writing routes	Backend orchestration
Scoring and targeted revision	Secret/token handling
Cover-image prompts and publishing checklist	Claims that content was published

Modes

Mode	Use when	Output
draft	Create a new platform draft	Purpose route, title options, body, tags, image prompt, checklist
score	Evaluate existing content	Hard-check result, 100-point score, top fixes
revise	Change part of a draft	Only revised section plus reason for changes
adapt	Convert one draft across platforms	Platform-specific versions preserving facts
package	Prepare for manual publishing	Copy-ready title/body/tags/image prompt/checklist

Default to draft when the user asks to write. Default to revise when the user points to a section or asks for a local change. Default to score when the user asks whether a draft is good enough.

Required First Decisions

1. Identify the writing purpose before generating or scoring content. See references/content-routes.md.
2. Identify target platform(s) and platform constraints. See references/platform-specs.md.
3. Identify voice/style source: user instruction, style-card, sample writing, or conservative generic voice. See references/style-systems.md.
4. Identify factual-risk level: concrete dates, numbers, names, claims, prices, legal/compliance statements, or medical/financial claims need verification or must be marked as uncertain.

If purpose or platform is missing, infer from the request when low risk. Ask one concise question only when the choice changes the output materially.

Workflow

Use references/workflow.md for the full execution sequence. Keep the visible output practical:

1. Purpose and platform route.
2. Draft or revision.
3. Hard checks.
4. Purpose-specific score.
5. Packaging for manual publishing.

Quality Bar

Use references/quality-check.md before returning final content. Every generated package must include:

• Copy-ready title and body.
• Platform-specific tags or keywords when relevant.
• Cover-image prompt with no text, no watermarks, no logos.
• Manual publishing checklist.
• Fact-risk notes when facts were not verified.

When revising, preserve unchanged facts and do not silently rewrite unrelated sections.

Optional Local Validation

If the user provides a draft file or asks for a check, run:

python3 scripts/check_content.py --platform xiaohongshu --file draft.md

The script only performs deterministic hard checks. Use the scoring rubric in references/quality-check.md for qualitative scoring.

References

File	Use for
references/workflow.md	Draft, score, revise, adapt, and package steps
references/content-routes.md	Writing-purpose routes and platform/content mapping
references/quality-check.md	Hard checks, scoring rubric, and revision rules
references/platform-specs.md	Platform format, length, title, and image constraints
references/style-systems.md	Voice systems and style-card guidance
references/publish-backends.md	Optional publishing handoff boundaries
references/runtime-compatibility.md	Runtime and dependency notes
assets/examples/	Style/platform card examples and quickstart sample