
from mnemonic11
Installs and configures a filesystem-backed mnemonic memory system for Claude-style agents, creating directory structure, CLAUDE.md hooks, and an initial contex
This skill configures a persistent, filesystem-based memory system (mnemonic) for Claude-style agents. It creates an idempotent directory structure, writes a user- and project-level CLAUDE.md memory protocol, initializes a git-backed memory store, and seeds an initial context memory so agents can search and capture memories across sessions.
Use this when you want an agent to maintain cross-session knowledge without external services — for onboarding projects, enabling proactive memory capture, or standardising memory namespaces across an organization. Run it once per project (safe to re-run).
Best suited for Claude-style agents or any LLM workflow that reads/writes local CLAUDE.md and filesystem memory (developers running local agent tooling).
Project setup skill group for scaffolding new projects and generating workflow files (CLAUDE.md, AGENTS.md, Plans.md). Contains only a SKILL.md with routing instructions to sub-skill doc.md files — no bundled scripts or referenced code files. Instructions are vague and reference doc.md files that aren't included, making the skill non-functional out of the box. Content is mostly in Japanese, limiting accessibility.
Skill is a meta/router that points to 5 sub-skills via doc.md references, but none of those files are included. Essentially a shell with no implementation. The SKILL.md frontmatter is valid and has good metadata. Security is high because there's nothing executable to audit — no scripts, no shell commands, no network calls. Architecture is below average due to missing referenced files and lack of separation of concerns.