create-post

---

name: create-post description: Creates a LinkedIn post for a team author. Use when the user asks to write, create, or make a LinkedIn post, or says "turn this into a post", "write a post about...", "[Author] wants to write about...". disable-model-invocation: true argument-hint: [Author] [topic or source material] allowed-tools: Read, Grep, Glob, Edit, Write, Bash(python *), Skill(validate-post *)

Create LinkedIn Post

Create a LinkedIn post for an author using the full 8-step pipeline.

Input

• $0 — author name (Seva, Kirill)
• Remaining arguments or conversation context — topic, source, voice input

Required files to load

Before starting, load:

1. Author DIP (required): Knowledge-Base/01-Identity-Profiles/[Author]-DIP.md

   • Mapping: Seva → Seva-Ustinov-DIP.md, Kirill → Kirill-Kasimskiy-DIP.md

2. Author Voice Vocabulary (required if the file exists): Knowledge-Base/01-Identity-Profiles/_Voice-Vocabulary/[Author]-Voice-Vocabulary.md

   • Contains the author's characteristic phrases from their transcripts and Telegram (NOT from published posts).
   • Used for humanization: in Step 4 when writing the body, prefer phrasing from the Voice Vocabulary over generic LinkedIn formulations.
   • If a Voice Vocabulary file for the author doesn't exist yet — proceed without it, but in the final report note "Voice Vocabulary for [Author] not created".

3. Post-Creation-Engine (required): Knowledge-Base/04-Post-Creation-Engine/Post-Creation-Engine.md

4. Algorithm-Intelligence (required): Knowledge-Base/02-LinkedIn-Algorithm/Algorithm-Intelligence.md

5. AI-Detection-Reference (required — for understanding 360Brew penalties): Knowledge-Base/02-LinkedIn-Algorithm/AI-Detection-Reference.md

6. Current-Trends (recommended): Knowledge-Base/03-Viral-Trends/Current-Trends.md

7. Hook-Pattern-Library (when picking a hook): Knowledge-Base/03-Viral-Trends/Hook-Pattern-Library.md

8. Benchmark-Validation-System (if the post will contain numbers): Knowledge-Base/04-Post-Creation-Engine/Benchmark-Validation-System.md

Pipeline

Step 0: Fresh Scan — refresh the DIP before working (MANDATORY)

CRITICAL PRINCIPLE: The DIP is built ONLY from transcripts and recorded speech. Posts are OUTPUT, not INPUT.

Before creating a post:

1. Open the author's DIP → check Last Updated and Last Fresh Scan
2. Find ALL new transcripts featuring the author after that date (sources — see the Source Map in CLAUDE.md)
3. Quickly scan each new transcript (5 min/file): new positions, stories, numbers, phrases
4. If something material is found → update the DIP before writing the post
5. Update the Last Fresh Scan date in the DIP

Run Steps 1–7 from Post-Creation-Engine.md. In brief:

Step 1: Identify the author and load the DIP

Extract from the DIP: Tone of Voice, Expertise Clusters, Stories Bank, Viral Angles, Format Preferences, "They NEVER" list.

Step 2: Extract the core message

One key idea. If multiple → propose a series. Classify by cluster and pillar from the DIP.

Step 3: Pick angle and format

• Angle: from DIP Viral Angles + Current-Trends.md
• Format: from Algorithm-Intelligence.md + DIP Format Preferences
• Check Anti-Patterns.md so you don't repeat dead patterns

Step 4: Write the post

• Hook: use Hook-Pattern-Library.md, adapt to the author's tone, 2–3 variants
• Body: short paragraphs, examples from DIP Stories Bank, first person (I/my)
• Voice Vocabulary integration: when writing the body, where the author already has a similar thought in the Voice Vocabulary (Sentence Starters, Idiomatic Phrases, Characteristic Adjectives/Verbs, Metaphors, Closing Patterns), prefer their phrasing over generic phrasings. Don't stuff in tics for the sake of count — point-swap generic phrases for authentic ones.
• CTA: specific, tied to the goal of the post (NOT "Agree?" / "Thoughts?" / 🔥)
• Length: sweet spot 1,200–1,800 characters
• What NOT to do (see detect-ai Levels 4F–4I): A-vs-B opener in the first 3 lines, false-vulnerability openers ("I'll be honest..." / "Hot take:"), "three lessons" numbered list in the closing, generic motivational CTAs, "one sentence per line + blank line between" carousel pattern longer than 5 lines in a row.

Step 5: Validation (built-in)

Run the basic check from Post-Creation-Engine.md Step 5. Full validation comes in the next step.

Step 6: Format output

1. Post text (copy-paste ready)
2. Alternative hooks (2–3 variants with notes)
3. Format recommendation (if visuals/carousel/video are needed — description)
4. Publish recommendation (day/time, coordination with other authors)
5. Source attribution (what came from which transcript/document)

Step 7: Save draft

• Path: Posted/[Author]/YYYY-MM-DD-[slug].md
• Format: per the template Knowledge-Base/04-Post-Creation-Engine/Post-Template.md
• Required: fill source_refs: in the frontmatter — list of ALL sources used

Step 8: VALIDATION

After saving the draft, run full validation:

/validate-post [path to the created file]

This runs:

• Fact-checking (/check-facts)
• Generic / repetition check (/check-generic)
• Source map (/check-sources)
• Final checklist (characters, format rotation, cluster, hook, anti-patterns, algorithm, mobile)

Show the user a summary validation report.

Step 9: Feedback (after any incoming feedback)

After EVERY piece of feedback from the author/team → update the DIP:

• "FEEDBACK LOG & LEARNINGS" section
• Format: Date | Post | Category | Feedback | Lesson Learned
• Categories: tone | accuracy | context | angle | format | strategy

Rules

• One post = one file. When reworking, overwrite the current file, do NOT create -v2, -v3
• First person. ALWAYS I/my in personal posts. NEVER we/our
• Fact-based. Every claim is backed by real data from sources
• Don't invent. Don't make up stories — use only the DIP Stories Bank and real sources
• Coordination. Check Posted/*/ for the past week — avoid topic overlap with other authors