Obsidian lets you capture messily. Codex helps you remember cleanly.
A platform-agnostic Codex + Obsidian workflow that turns user-captured inspiration, content references, project notes, and long creative history into structured, searchable, task-specific memory.
Capture messily. Organize periodically. Execute with focused context.
This is a manual workflow template, not a background watcher, scraping tool, content automation system, platform login automation project, or media downloader.
This workflow is designed for user-captured inspiration, not automated scraping. You save your own notes, links, screenshots, recordings, observations, and examples into Obsidian. During a manual review session, Codex can review those materials, classify them, extract reusable patterns, update memory files, and support task-specific execution.
This template is for creators, consultants, founders, marketers, researchers, educators, and small teams who collect ideas across many tools but do not want those materials to become a messy archive.
You can use it for Xiaohongshu, TikTok, Instagram, YouTube, LinkedIn, X, newsletters, blogs, podcasts, personal brands, consulting content, product marketing, community operations, or any recurring creative workflow.
Obsidian is the durable memory layer.
Codex is the manual review partner and task-specific execution partner.
Task Packs define what Codex should read, write, check, and stop after doing.
Patterns are reusable knowledge extracted from messy inspiration.
Messy capture across platforms and creative workflows
↓
Obsidian Inbox
↓
Manual Codex review
↓
Classification, tagging, scoring, pattern extraction
↓
Platform-agnostic patterns
↓
Short context indexes, Task Packs, and project memory
↓
Task-specific Codex execution
↓
Content, thumbnails, carousels, videos, newsletters, blogs, lead tracking, QA, weekly planning
↓
Error correction and memory update
Screenshots, links, recordings, and half-formed ideas do not stay as messy reference piles. During review, Codex can classify them, score adaptation value, check copying risk, and extract reusable patterns.
Codex should not read the entire vault by default.
For each task, Codex should usually read only:
Obsidian_Template/00_Index/operating_context_index.md- the current Task Pack
- relevant pattern files
- a small number of necessary project files
Long-term memory lives in Obsidian.
Current execution uses short context.
When Codex makes a recurring mistake, the fix should not stay buried in chat history. Record it in an error correction template, update the relevant Task Pack or rulebook, and make the next run safer.
Inbox: messy capture layer for raw notes, links, screenshots, recordings, and observations.Inspiration: classified reference material organized by creative format or strategy type.Patterns: extracted reusable rules, formulas, and workflow patterns.Memory: project rules, decisions, long-term context, and working agreements.TaskPacks: read/write/check/stop boundaries for Codex tasks.
This template is not tied to one platform.
Use it for social media, creator workflows, consulting content, newsletters, product marketing, community operations, or any repeatable creative process.
The workflow is designed around user-captured inspiration. It does not provide scraping tools, platform automation, login automation, or content downloading scripts.
Respect platform terms, copyright, creator rights, and privacy.
Capture like a P.
Organize like a J.
Execute with Codex.
Remember with Obsidian.
- Copy
Obsidian_Template/into your Obsidian vault. - Read
docs/quickstart.md. - Add one raw idea, link, or screenshot note to
Inbox/. - Ask Codex: "Run TaskPack_Inbox_Review.md on the latest inbox items."
- Let Codex classify the item, extract a pattern, and update the memory files.
This public repository contains only templates and sanitized examples.
Do not include:
- private brand, client, or project names
- private messages
- CRM data
- student, tutor, customer, or team member information
- account data or real business metrics
- raw screenshots with personal data
- unpublished strategy
- private media files
- local file paths
- login details
- API keys or access tokens
- private Obsidian vault content
- unaudited skills or internal automation scripts
Use only:
- Mock Brand
- Mock Project
- Mock Screenshot Description
- Mock Content Example
- Mock TaskPack Run
Code is licensed under the MIT License.
Templates, documentation, and examples are licensed under CC BY-NC 4.0 by default. See CONTENT_LICENSE.md.