Stats
Actions
Tags
From hipocampo
Scaffold a hipocampo memory vault into the current project — creates brain.config.toml and the docs/brain skeleton (knowledge/index, capture.md, context-budget.md, note templates). Use when setting up agent memory in a repo that doesn't have a vault yet, or when the user says "init the brain", "set up the vault", "brain-init".
How this skill is triggered — by the user, by Claude, or both
Slash command
/hipocampo:brain-initThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Create a durable, git-versioned, human-gated memory vault in this repo. Idempotent:
Create a durable, git-versioned, human-gated memory vault in this repo. Idempotent:
never overwrite an existing .md.
Prerequisites: ability to run git and Python (python3, or python on
Windows). In a restricted sandbox that blocks writes/process execution, these
skills cannot scaffold — say so instead of half-running.
brain.config.toml already exists, read it and
reuse its answers. Otherwise gather, via ONE AskUserQuestion (4 questions —
this is the whole interview; don't ask more):
brain-router-init detects and confirms; new/greenfield → it asks the
stack questions instead)conservative (decisions/contracts only) /
balanced (+ lessons and sources, recommended) / aggressive (+ any new
concept)en or pt-BR)
Vault root (docs/brain) and initial areas are sensible defaults — state
them in your summary as confirmable, don't ask.brain.config.toml at the repo root from
${CLAUDE_PLUGIN_ROOT}/brain.config.example.toml, filling in the answers.
Record the interview for the other skills as plain keys:
project_mode = "existing"|"greenfield" and team = true|false.
Don't overwrite an existing config — diff and ask first.${CLAUDE_PLUGIN_ROOT}/templates/vault/<language>/
into <vault_root>/, skipping any file that already exists. Preserve the empty
dirs (insights/, raw/sources/, knowledge/_inbox/, specs/, adrs/) —
they ship with .gitkeep; a recursive copy keeps them. Then replace the placeholders in every copied file: {{DATE}} with today's
date (ISO YYYY-MM-DD) and {{CAPTURE_LEVEL}} with the chosen level (use
the locale's word: en conservative/balanced/aggressive, pt-BR
conservador/equilibrado/agressivo).{{DATE}}/{{CAPTURE_LEVEL}} (fail and re-render if
any remain), then run vault_sync. The hipocampo package isn't vendored yet
(that's brain-scripts-init), so run it from the kit:
PYTHONPATH=${CLAUDE_PLUGIN_ROOT} python3 -m hipocampo.validators.vault_sync
(use python if python3 isn't on PATH). A fresh vault must pass (index
present, no FAILs).capture.md (how chat becomes durable notes) and context-budget.md
(index-first reads).brain-router-init (the AGENTS.md router) and
brain-scripts-init (vendor the validators + git hooks).npx claudepluginhub ferraz2000/hipocampo-memory --plugin hipocampoProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.