Agent-Skills.md

Agent Skills: Relearning Content Creator

Creates journal entries or project pages for a personal knowledge site. Use when the user wants to write, publish, or add content - journals, projects, or articles about cognitive engineering, productivity systems, or tool-driven growth.

UncategorizedID: kedbin/relearning-flow/relearning-content

Repository

kedbin/relearning-flow
kedbin
2

Install this agent skill to your local

pnpm dlx add-skill https://github.com/kedbin/relearning-flow/tree/HEAD/skills/relearning-content

Skill Files

Browse the full folder contents for relearning-content.

Download Skill

Loading file tree…

skills/relearning-content/SKILL.md

Skill Metadata

Name
relearning-content
Description
Creates journal entries or project pages for a personal knowledge site. Use when the user wants to write, publish, or add content - journals, projects, or articles about cognitive engineering, productivity systems, or tool-driven growth.

Relearning Content Creator

Creates structured journal entries or project pages following a cognitive engineering philosophy and Astro content schema.

The Philosophy

Core Mission: Apply enterprise-grade engineering discipline to the messy reality of being human. Debug sleep, focus, and decision-making as if they were mission-critical infrastructure.

The Lens: Every human problem is reframed as a systems engineering problem. We don't moralize; we diagnose. We don't motivate; we architect.

The Promise: No hype. Just honest metrics. If something failed, log it. If a belief was wrong, document the update.

When to use this skill

USE THIS SKILL when the user:

  • Wants to create a new journal entry
  • Wants to add a new project page
  • Says "new journal", "new entry", "write about [topic]"
  • Provides content/ideas and wants them formatted for the site
  • Wants to document a project, tool, or system they've built

IMPORTANT: This skill creates content files. After content is finalized:

  1. Use create-script skill to condense + add paralinguistic tags → saves .txt
  2. Use voiceover skill on the .txt file → generates .mp3 + deploys + pushes

Workflow Architecture (CRITICAL)

Your role as the main agent is REVIEWER, not drafter.

The google-search subagent handles:

  • Research (finding sources, opposing views)
  • Drafting (writing the complete journal entry)
  • Citation integration

You handle:

  • Repository sync and file management
  • Tone/consistency review against standards
  • Iteration requests if draft doesn't match voice
  • Final file creation and audio pipeline
User provides topic/content
        ↓
Step 0: Git pull + determine entry number
        ↓
Step 1: Spawn google-search subagent to DRAFT
        - Subagent researches topic
        - Subagent writes complete journal entry
        - Subagent returns full markdown
        ↓
Step 2: YOU review for tone/consistency
        - Does it match the voice?
        - Engineering metaphors present?
        - Fallacy → Model → Protocol structure?
        - Memorable one-liner ending?
        ↓
Step 3: If lacking, send back to subagent with feedback
        ↓
Step 4: Save final draft to entry-XXX.md
        ↓
Step 5: Present to user for approval
        ↓
Step 6: Audio pipeline (create-script → voiceover)

How to Execute This Skill

Step 0: Sync Repository (ALWAYS DO THIS FIRST)

cd ~/projects/your-site && git pull origin main
ls src/content/journal/
date +%Y-%m-%d  # Get today's date for the entry

Determine the next entry number (e.g., if entry-013.md exists, next is entry-014.md).

CRITICAL: Use TODAY'S DATE as the publish date. Run date +%Y-%m-%d to get the current date. Do NOT use the date from the user's notes - that is their draft date, not the publish date.

Step 1: Delegate Drafting to google-search Subagent

Spawn the subagent with the user's content and request a complete draft:

Task(subagent_type="google-search", prompt="Draft a journal entry on the following topic:

[USER'S CONTENT/IDEAS HERE]

Requirements:
1. Research the topic thoroughly - find relevant studies, frameworks, and opposing viewpoints
2. Write a complete journal entry following the style guide (see your instructions)
3. Use the Fallacy → Model → Protocol structure
4. Include at least 3 citations with proper references
5. End with a memorable one-liner
6. Return the complete markdown file ready for publication

Entry number: entry-XXX
Date: YYYY-MM-DD")

Step 2: Review the Draft for Tone/Consistency

When the subagent returns, check:

Voice Checklist:

  • [ ] Title follows "The [Technical Noun]: [Subtitle]" pattern
  • [ ] Engineering metaphors used throughout (not generic self-help language)
  • [ ] Problems framed as bugs/inefficiencies, solutions as protocols/patches
  • [ ] Academic rigor - citations present with author, year
  • [ ] Summary is systems-framed, 1-2 sentences
  • [ ] At least 3 highlights with metrics/sources
  • [ ] Opening hook is personal/specific, not generic
  • [ ] Fallacy section identifies legacy thinking
  • [ ] Model section cites named frameworks with authors
  • [ ] Protocol section has numbered phases
  • [ ] Ends with memorable, quotable one-liner
  • [ ] References section complete

Red Flags (send back for revision):

  • Generic motivational language ("unlock your potential", "achieve your dreams")
  • Missing citations or vague claims
  • No engineering/technical metaphors
  • Protocol section too abstract (needs concrete actions)
  • Weak or missing one-liner ending

Step 3: Iterate if Needed

If the draft doesn't match the voice, spawn the subagent again with specific feedback:

Task(subagent_type="google-search", prompt="Revise this draft:

[PASTE DRAFT HERE]

Issues to fix:
1. [Specific issue - e.g., 'Opening hook is too generic, needs a specific incident']
2. [Specific issue - e.g., 'Missing engineering metaphor for willpower concept']
3. [Specific issue - e.g., 'Protocol section needs concrete metrics']

Return the revised complete markdown.")

Step 4: Save the Final Draft

Once the draft passes review, save it:

# Write to entry file
~/projects/your-site/src/content/journal/entry-XXX.md

Step 5: Present to User

Show the user:

  • The filename created
  • A summary of the content
  • The highlights/key takeaways
  • Ask for any revisions

Step 6: Audio Pipeline (After User Confirms)

Once user approves, execute the two-step audio pipeline:

Step 6a: Create Voiceover Script

Use the create-script skill to condense and add paralinguistic tags.

Step 6b: Generate Audio

Run the voiceover command and only verify it started (do not poll for progress):

cd ~/projects/chatterbox && nohup uv run python archive/voiceover_script.py \
  -i archive/entry-XXX.txt \
  -o archive/entry-XXX.mp3 \
  --entry entry-XXX \
  --push > voiceover.log 2>&1 &

Then verify it started:

sleep 5 && head -10 ~/projects/chatterbox/voiceover.log

DO NOT poll for progress repeatedly. Trust that the script will complete and push. The user will receive a desktop notification when done.

Tell the user:

  • Voiceover generation launched in background
  • They will receive a desktop notification when complete
  • Can monitor with: tail -f ~/projects/chatterbox/voiceover.log

Project Location

Repository Path: ~/projects/your-site (configure to your setup)

  • Journal entries: src/content/journal/entry-XXX.md
  • Project pages: src/content/projects/[slug].md

Content Schemas

Journal Entry Schema

---
title: "The [Metaphor]: [Subtitle with Engineering Framing]"
date: "YYYY-MM-DD"
summary: "[1-2 sentence hook with systems/engineering lens]"
status: "Published"
category: "Relearn [Life|Engineering|Work] / [Subcategory]"
highlights:
  - "Key Takeaway 1: [Actionable insight]"
  - "Key Takeaway 2: [Framework or model]"
  - "Key Takeaway 3: [Protocol or implementation]"
audioUrl: "/audio/entry-XXX.mp3"
---

Project Page Schema

---
title: "[Project Name]: [Subtitle]"
date: "YYYY-MM-DD"
description: "[1-2 sentence description]"
repoUrl: "https://github.com/yourusername/[repo]"
demoUrl: "[URL]"
techStack: ["Tech1", "Tech2", "Tech3"]
audioUrl: "/audio/[slug].mp3"
---

Voice Reference (For Your Review)

Good Examples (Match This Tone)

Titles:

  • "The Physics of Productivity: Mastering the Input/Output Ratio"
  • "The Asymptote of Effort: Overcoming the Iron Law of Diminishing Returns"
  • "Memoization: The Architecture of Cognitive Caching"

Summaries:

  • "A system running at 100% utilization with 0% throughput is not 'dedicated'—it is broken."
  • "Most human exhaustion comes from re-computing solved problems."

One-Liners:

  • "Stop watching reality. Start predicting it."
  • "Stop calculating. Start retrieving."
  • "Stop acquiring tools. Start becoming them."

Bad Examples (Reject This Tone)

  • "Unlock your full potential with these 5 simple steps"
  • "The secret to success is believing in yourself"
  • "Transform your life with the power of positive thinking"

Quick Reference: Technical Metaphors

| Human Concept | Engineering Metaphor | |---------------|---------------------| | Decision fatigue | Memory leak, garbage collection failure | | Willpower | Battery charge, finite resource pool | | Habits | Compiled routines, cached functions | | Procrastination | System deadlock, CPU thrashing | | Attention | Single-core processor, context switching | | Goals | Function signatures, API contracts | | Feedback | Control loops, negative feedback systems | | Learning | Compiling, updating dependencies | | Forgetting | Cache invalidation, memory volatility | | Burnout | Thermal throttling, system overload |

Astro Markdown Rules

  • NO markdown tables (use bullet lists)
  • NO code blocks with language hints
  • ASCII-safe frontmatter (spell out special characters)
  • NO raw HTML
  • NO footnotes (use [1] citation style with References section)

Important Reminders

  • Delegate drafting to google-search subagent
  • Your job is review for tone/consistency
  • Iterate if needed - send back with specific feedback
  • Only verify voiceover started - don't poll for progress
  • Trust the pipeline - script handles deploy + push + notification

The Workflow:

Git Pull → Subagent Drafts → You Review → Iterate if Needed → Save → User Confirms → create-script → voiceover (fire and forget) → Done!
Relearning Content Creator Skill | Agent Skills