Content Pipeline

Do NOT load this skill for voice/tone questions (use personal-identity), engineering decisions (use engineering-knowledge), contact lookups (use network), or goal tracking (use operations-memory).

Files in This Module

All paths relative to ~/brain/:

Setup check: If any of these directories or files is missing, inform the user and suggest creating them. For entity files, use the templates in skills/content-pipeline/assets/templates/entity-templates/.

Behavioral Rules

These rules operationalize AGENT.md Rules #1 and #10 for this module. They are non-negotiable for all content pipeline tasks.

Rule CP-1 — Ideas Pipeline Review

When the user asks to review their ideas pipeline or asks "what should I write about":

1. Glob ~/brain/content/ideas/*.md
2. Read frontmatter only from each file: id, status, score_total, priority, target_platform, tags
3. Filter to ideas with status raw or scored
4. Sort by score_total descending, then by priority (high > medium > low)
5. Return a structured list of the top 5 actionable ideas
6. Never load the full body content of idea files during pipeline review

Rule CP-2 — Content Gap Analysis

When the user asks about content gaps or what topics need coverage:

1. Glob ~/brain/content/posts/*.md
2. Filter to posts where date_published is within the last 30 days
3. Count posts per pillar from frontmatter
4. Load ~/brain/brand/pillars.yaml to get the full list of defined pillars
5. Compare: any pillar with zero posts in the last 30 days is a gap
6. Present gaps ranked by staleness (longest since last post first)

Rule CP-3 — Duplicate Check

Before creating any new content idea:

1. Glob ~/brain/content/ideas/*.md and ~/brain/content/posts/*.md
2. Extract the title field from frontmatter of each file
3. Compare the proposed idea title against existing titles for semantic overlap
4. If a likely duplicate exists: warn the user and show the matching file(s)
5. Only proceed with creation after the user confirms it is not a duplicate

Rule CP-4 — Scoring Workflow

When the user asks to score an idea:

1. Load ~/brain/identity/voice.md and ~/brain/brand/pillars.yaml as scoring inputs
2. Evaluate the idea across 5 dimensions, each scored 1–5:
   • relevance — how well it aligns with defined content pillars
   • uniqueness — how differentiated it is from existing published content
   • audience_fit — how well it serves the target audience from brand positioning
   • effort — estimated effort to produce (lower score = easier = better)
   • total — sum of the four scores above
3. Update the idea file frontmatter with all five scores
4. Update status from raw to scored
5. Present the scorecard to the user with brief justification for each dimension

Rule CP-5 — Bookmark Capture

When the user asks to save a bookmark, URL, or resource:

1. Create a new file in ~/brain/knowledge/bookmarks/ using the bookmark entity template
2. Generate the id as bm-YYYY-NNN where YYYY is the current year and NNN is the next sequential number
3. Fill in all required frontmatter fields: date_saved, type, url, author, status (default: unread)
4. Ask the user for a key takeaway if they haven't provided one
5. Never batch-load all bookmark files — only read individual bookmarks when specifically referenced

Rule CP-6 — Draft Workflow

When the user asks to draft content:

1. Always load personal-identity first per AGENT.md Rule #10 — this ensures voice.md and anti-patterns.md are applied
2. Determine the target platform from the user's request
3. Use the corresponding platform template from skills/content-pipeline/assets/templates/:
   • Blog → blog-template.md
   • LinkedIn → linkedin-template.md
   • Newsletter → newsletter-template.md
   • Twitter/X thread → thread-template.md
4. If an idea file exists for this content, load it to pull in the hook and notes
5. Write the draft into ~/brain/content/drafts/ with a descriptive filename
6. Run anti-patterns check before presenting the draft to the user

Observation Masking

When reviewing ideas pipeline: glob content/ideas/*.md, load frontmatter only. For scoring: filter status raw or scored. For duplicates: compare title field only. Return structured list of top 5 actionable ideas.

When reviewing posts: glob content/posts/*.md, filter by date_published, load frontmatter only. Never return full body content during pipeline review.

When browsing bookmarks: glob knowledge/bookmarks/*.md, load frontmatter only. Only load full bookmark content when the user asks about a specific bookmark by name or topic.

Platform Templates

These files live in skills/content-pipeline/assets/templates/ and define the structure for each content platform:

Entity Templates

These files live in skills/content-pipeline/assets/templates/entity-templates/ and define the frontmatter schema for each entity type:

Version Notes

• v1.0.0 — Phase 2A: initial content-pipeline sub-skill. Covers content ideas, posts, drafts, bookmarks, and calendar. Entity directories with observation masking. Behavioral rules CP-1 through CP-6 implement AGENT.md Rules #1 and #10 at the module level.