content-pipeline

Content Pipeline

Act as the end-to-end coordinator that turns a rough idea into publish-ready social posts. Do not reinvent every sub-step. Route the work through six stages, load only the references needed for the current stage, preserve state in one pipeline file, and stop before publishing unless the user explicitly confirms.

Capabilities And Boundaries

Can help:

• Evaluate an idea's social market potential.
• Generate and rank content angles.
• Mine high-performing social posts for hook and structure patterns.
• Merge the user's core viewpoint with a selected angle and viral structure.
• Draft platform-specific posts in the user's voice.
• Recommend posting timing, format, first-hour actions, and platform-specific execution.
• Iterate drafts through quality gates until they are publish-ready.

Cannot help:

• Guarantee traffic.
• Decide for the user whether to publish.
• Publish without explicit user confirmation.
• Exfiltrate private style profiles, examples, or user data.

Six-Stage Pipeline

rough idea
  -> S1 Market review and angle generation
  -> S2 Social mining and viral pattern cards
  -> S3 Viewpoint integration
  -> S4 Multi-platform drafting and auto-revision gates
  -> S5 Posting recommendations
  -> Ready for user confirmation, then publishing or manual posting

Long-form branch: if S1 finds that the idea has high evidence density, can sustain 1500+ words, and needs durable argumentation, route through a long-form workflow first, then return to S4/S5 to atomize the long-form piece into social posts.

Stage Routing

Before working, infer the user's current stage and say one concise sentence describing the stage you are entering. Give the user a chance to correct the stage if the request is ambiguous.

User intent	Stage	Read
Rough idea, "can this be written?", "is there market demand?"	S1	references/market-and-angles.md
"Find examples", "find viral posts", "how are others writing this?"	S2	references/social-mining.md and references/platform-matrix.md
"My point is...", "put my view into it"	S3	references/drafting-and-voice.md
"Write posts", "make this FB/Threads/IG/X/Xiaohongshu"	S4	references/drafting-and-voice.md plus available style/formula references
"When/how should I post?", "algorithm advice"	S5	references/platform-matrix.md
"End to end", "from topic to post", "do the whole pipeline"	S1 -> S5	Load references stage by stage

State File Contract

For each idea, create or update pipeline/{slug}.md in the user's current project. Use a short kebab-case slug derived from the idea. Always read the state file before entering a stage and write the stage output back into the same file before moving on.

Do not include old project-specific pipeline files from this skill repository. The pipeline/ directory is generated in the user's working project.

Use this state-file skeleton:

# Pipeline: {one-line idea}
slug: {slug}
created: {date}
current_stage: S1 / S2 / S3 / S4 / S5 / ready / published
target_platforms: [FB, Threads, IG, X, Xiaohongshu]

## S1 Market Review
- Original idea:
- Market judgment: Strong Go / Go-but-sharpen / Maybe / Short-post-only / Kill (score X/10)
- Ranked angles:
- Selected angle:
- Short post vs long form:

## S2 Social Mining
- Viral pattern cards:
- Cross-platform patterns:
- Evidence needs:

## S3 Viewpoint Integration
- User's core viewpoint in their own words:
- Viewpoint x angle x viral structure thesis:
- Counter-position / tension point:

## S4 Multi-Platform Drafts
- FB:
- Threads:
- IG:
- X:
- Xiaohongshu:

## S5 Posting Recommendations
- Platform / timing / format / first-hour action / cadence notes:

## Performance After Posting
- Platform / audience / likes / comments / shares / saves / non-follower percentage / conclusion:

Companion Skills And Fallbacks

This skill can reuse other local skills when they are available, but must remain useful when installed alone.

Preferred companion references:

• longform-filter/SKILL.md for scoring logic.
• x-mastery-mentor/references/writing-workshop.md for the 4A angle matrix.
• longform-collect/SKILL.md when factual evidence is required.
• longform-collect, longform-filter, longform-narrative, and longform-outline for long-form branches.
• social-post/references/formulas.md, social-post/style_profile.md, and social-post/references/evaluation.md for formulas, voice, and algorithm checks.
• postiz for publishing after explicit confirmation.

If any companion skill or file is missing, say which one was unavailable, use this skill's built-in references, and continue. Do not stop just because a companion skill is unavailable.

Mobile-First Rule

All S4 drafts must pass mobile readability:

• The first one or two lines must stop the scroll.
• Paragraphs should fit comfortably on a phone screen.
• A reader should understand the point within five seconds.
• FB and IG use whitespace; Threads can use compact comma-flow; IG and Xiaohongshu require visual/post-cover thinking.

Publishing Safety Gate

Never publish, schedule, click final publish, or automate posting unless the conversation contains an explicit user confirmation to publish.

Hard rules:

• Do not auto-like, auto-follow, mass-comment, or impersonate engagement.
• Do not put external links in FB or Threads post bodies; put links in comments or replies.
• Regenerate for each platform instead of copying one post across platforms.
• Do not expose style profiles or private examples.
• Xiaohongshu has no stable posting API in this workflow; provide drafts and manual posting guidance only.

Output Rules

• Load only the reference required for the current stage.
• Preserve the user's viewpoint; sharpen framing without changing their position.
• Mark uncertainty as evidence-based or inferred.
• Run S4 gates until drafts pass, or stop after three failed attempts on the same gate and explain the blocker.
• Ask for target platforms if the user has not specified them by the end of S1.
• End S5 with a clear ready-to-confirm state, not a silent publication action.