obsidian-gh-knowledge
Obsidian GH Knowledge (CLI-first)
TL;DR
- Goal: Bootstrap and operate the Obsidian vault using the safest and most optimal method depending on the environment context.
- Execution Order: 0. Bootstrap a local clone after repo confirmation -> 1. Local Obsidian CLI -> 2. Local fallback -> 3. GitHub API fallback.
- Agent Rules: Agents must read before write, include a
## TL;DR, and must use Mermaid diagrams for visual explanations. - Project Scoping: When working on a specific project (cmux, trends, etc.), stay within that project's folder under
5️⃣-Projects/GitHub/<project>/. See "Project Scoping (CRITICAL)" section below. - Local Wrapper: Use
scripts/local_obsidian_knowledge.pyfor repeatable local macOS workflows that combine Obsidian CLI operations with this vault's project rules and git sync. - Raw Materials: If the vault mounts
raw/as a Git submodule, treat it as source-input storage rather than curated note space. - Inbox Split:
raw/inboxis the default intake lane for external source material.0️⃣-Inboxis curated staging for notes that already contain synthesis and still need routing. - Health Default: Use
simplify-reviewfirst when the vault feels hard to read or hard to trust; it reconciles full-vault Obsidian counts with active-scope graph checks and overview readability audits. - Git Sync: For note updates, prefer pull/rebase before push, never force-push
main, and record the sync outcome in today's daily log when the task is complete. - Git Bootstrap: Local bootstrap must configure Git and the
raw/submodule even when the Obsidian CLI is unavailable. Keep the baseline hook-free and clean up stalecore.hooksPathconfig during bootstrap.
Execution Mode Flowchart
More from karlorz/agent-skills
simplify
Review and simplify recently changed code for reuse, clarity, and efficiency while preserving behavior. Use when the user asks to simplify, refine, polish, clean up, or make code clearer, or after finishing a logical chunk of implementation that should be tightened before commit.
41autopilot
Install or refresh a repo-local Claude autopilot bundle into the current repository or an explicitly provided target repository. Use when asked to set up repo-local autopilot hooks, install autopilot into `.claude/`, refresh a repo-local autopilot bundle to the latest design, or reset a repo's local autopilot workflow.
7ralph-loop
Keep Codex auto-continuing on the same bounded task until a completion signal appears in the assistant output or a maximum iteration count is reached. Use when the user asks for a Ralph Loop, asks Codex to keep going until done, or wants to prevent premature stopping on a task that should stay in the current workspace.
2