PARA Second Brain

**Quick judgment** A strong, well-documented skill that upgrades Clawdbot’s memory into a structured PARA-based second brain with broad semantic search. Best suited to users willing to maintain a disciplined note-taking workflow and comfortable with basic filesystem operations. **What it does well** - Turns the workspace into a coherent knowledge system: `MEMORY.md` for curated long-term memory, `memory/YYYY-MM-DD.md` for daily logs, and `notes/` organized by PARA (projects, areas, resources, archive). - Uses a **symlink trick** so memory_search can index the entire PARA notes directory, effectively enabling semantic search across all your notes, not just MEMORY.md and daily logs. - Adds optional **session transcript indexing** via config so conversations become part of the searchable knowledge base. - Defines a **memory flush protocol** tied to `session_status` context thresholds, giving the agent clear rules for when to offload important information to files to avoid losing it during compaction. - Includes rich guidance (templates, decision trees, quality gates) to keep notes high-quality and future-friendly, not just a dumping ground. **Key risks / limitations** - **Symlink / OS issues:** The approach assumes a filesystem that supports symbolic links and paths similar to the examples (UNIX-like). Windows users or non-standard environments may need adjustments; misconfigured symlinks can make notes non-searchable or accidentally link the wrong directory. - **Implementation coupling:** It relies on current behavior of `memory_search` (what it indexes, how it treats the `memory/` directory and `sessions`). Changes in the underlying tooling could reduce effectiveness until configs are updated. - **User discipline required:** The system is only as good as the notes maintained. If daily logs, PARA folders, and MEMORY.md aren’t kept up to date, the “second brain” degrades quickly. - **Privacy / data surface:** Indexing session transcripts and broad local notes makes more information easily retrievable. In shared environments or on insecure machines, this can increase the impact of any unauthorized access. - **Version mismatch nuance:** The page text references features labeled as v2.0 while the provided metadata lists skill version 1.0.0, suggesting docs may be ahead of or not perfectly aligned with the published version. Users should verify that described features (especially transcript indexing and symlink behavior) match their installed version. **Recommended scenarios** - You want **local, zero-additional-cost** long-term memory for Clawdbot and are comfortable running shell commands and editing config files. - You run **multi-session, multi-project** workflows and need the agent to reliably recall prior decisions, tasks, and discussions. - You already use or are willing to adopt **PARA or similar PKM structures**, and want the agent to navigate that structure semantically. - You value **process clarity**: explicit rules on what to save, where to save it, and how to prevent context loss. Not ideal if you want a fully managed, cloud-hosted knowledge base with minimal setup, or if you’re unwilling to maintain consistent note-taking and curation habits.