portal-engram-memory

Persists development decisions and session context across Claude Code sessions using a local Engram SQLite database.

Updated Jul 27, 2026
One-click install
npx skills add https://github.com/ArthurZizumbo/karisma-data --skill portal-engram-memory-arthurzizumbo
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: portal-engram-memory
Source: https://github.com/ArthurZizumbo/karisma-data/tree/main/.claude/skills/portal-engram-memory
Command: npx skills add https://github.com/ArthurZizumbo/karisma-data --skill portal-engram-memory-arthurzizumbo

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Development teams lose decisions, rubric changes, and agreed trade-offs between coding sessions, forcing repeated re-explanation of project context to the AI assistant. ## Core Features & Use Cases - Session Memory Persistence: Save team decisions, rubric absorptions, agreed degradations, and fulfilled gates to a local Engram SQLite database with FTS5 search. - Session Lifecycle Workflow: Confirm project context at session start with mem_current_project and mem_search, save decisions as they occur with mem_save, and close with mem_session_summary. - Secret-Safe Guardrails: Enforces hard rules against storing credentials, JWTs, API keys, or participant PII, keeping Engram strictly outside production runtime code. - Use Case: A team absorbs a new grading rubric mid-sprint; the skill records the criterion-to-weight mapping and story point recalculation so the next session on any laptop starts with that context already loaded. ## Quick Start Ask the assistant to confirm the current Engram project and search memory for prior decisions relevant to today's task before starting work.

Frequently Asked Questions about portal-engram-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I persist decisions between Claude Code sessions?

Use the mem_save MCP tool to record decisions as they happen, including what, why, where, and learned fields. At session start, call mem_current_project and mem_search to recover prior context, and close with mem_session_summary.

What is Engram and how does it store memory?

Engram is a local-only Go binary that stores development memory in a SQLite database with FTS5 full-text search, located outside the repository. Cloud sync is opt-in and disabled for this project; team sharing happens via reviewed git chunks.

Can Engram be used for the production agent's conversational memory?

No. Runtime agent memory belongs to the ADK SessionService with PostgreSQL, and catalog retrieval uses pgvector. Engram is strictly a dev-time IDE tool and must never connect to FastAPI routers, ADK tools, or production code.

What data should never be saved to Engram memory?

Never store .env values, JWT secrets, API keys, Bearer tokens, passwords, raw user prompts, or real participant emails from UX research. Memory stores the why behind decisions, not secrets or data already versioned in git.

How do I sync Engram memory across team laptops?

Run engram sync to export chunks, review and redact them, then commit only chunks under docs/engram/ after a teammate review. On another laptop, pull and run engram sync --import; local .engram state stays git-ignored.