memory-export

name	memory-export
description	Export all cognitive memory to a portable file for porting Alex to other AI work surfaces
tier	extended
applyTo	**/*export*memory*,**/*memory*export*,**/*port*alex*

Memory Export Skill

Export the complete set of stored memories, user profile, preferences, and learned context into a single portable file. Designed for porting Alex's knowledge to other AI work surfaces (Claude Code, ChatGPT, etc.).

Purpose

Alex accumulates knowledge across multiple memory layers. This skill collects all layers into one structured, copy-pasteable document that can seed a new AI session elsewhere.

Memory Sources to Collect

Gather from ALL of these sources, in order:

Source	Location	Contains
User Profile	AI-Memory/user-profile.json	Identity, career, credentials, preferences, technologies (cloud-synced)
User Memory	/memories/ (Copilot Memory files)	VS Code workflow tips, tool workarounds, writing rules
Session Memory	/memories/session/	Current conversation context (if any)
Repository Memory	/memories/repo/	Codebase-specific facts stored by Copilot
AI-Memory Profile	AI-Memory/profile.md	Narrative profile and expertise (cloud-synced)
Active Context	.github/copilot-instructions.md Active Context section	Current persona, phase, mode, focus trifectas
Repository Memories	Repository memories in system context	Facts stored from prior agent interactions

Export Format

Output as a single Markdown document inside a code block. Use this exact structure:

# Alex Memory Export
# Exported: {YYYY-MM-DD}
# Sources: {list sources that had content}

## Instructions
[YYYY-MM-DD] - {verbatim rules the user set}

## Identity
[YYYY-MM-DD] - {personal facts}

## Career
[YYYY-MM-DD] - {roles, companies, skills}

## Projects
[YYYY-MM-DD] - {project name — what it does, status, key decisions}

## Preferences
[YYYY-MM-DD] - {opinions, working style, tool preferences}

## Architecture Context
[YYYY-MM-DD] - {Alex version, trifecta counts, active focus, current phase}

Format Rules

1. One entry per line, sorted oldest-first within each section
2. Date prefix: [YYYY-MM-DD] or [YYYY-MM] if only month known, or [unknown] if undated
3. Preserve the user's words verbatim for instructions and preferences — do not paraphrase
4. Deduplicate: if the same fact appears in multiple sources, emit it once with the most specific date
5. Wrap entire output in a single code block for easy copying
6. After the code block, state completeness: "Complete set" or "Partial — {what's missing and why}"

Section Guidelines

Instructions: Only rules the user explicitly asked to be followed. Tone, format, style, "always do X", "never do Y", behavioral corrections. Do NOT include architecture rules from copilot-instructions.md — those travel with the repo.

Identity: Name, age, birthday, location, origin, education, family, languages, personal interests. Source primarily from user-profile.json.

Career: Current and past roles, companies, impact metrics, expertise areas. ONE entry per role/highlight preferred.

Projects: ONE entry per project. Include: what it does, current status, key decisions. Use project name as first words. Merge information from multiple sources about the same project.

Preferences: Working style, learning style, tool preferences, diagram preferences, model preferences, publishing goals. Broadly applicable opinions.

Architecture Context: Current Alex version, phase, mode, trifecta counts, focus areas. This section helps the receiving surface understand where things stand.

Portability Notes

The export is designed to be pasted into:

• Claude Code (CLAUDE.md or session context)
• ChatGPT (custom instructions or project knowledge)
• Other Copilot workspaces (as seed context)
• Any LLM (as system prompt supplement)

The receiving surface won't have Alex's architecture, so:

• Omit references to specific file paths within the Alex architecture
• Omit synapse connections, trifecta internals, and cognitive protocols
• Focus on what the user wants the AI to know and do, not how Alex implements it
• Keep technical project details that help any AI assist the user

Anti-Patterns

Don't	Why
Include architecture internals (synapses, hooks, muscles)	Not portable — specific to Alex
Paraphrase user instructions	Loses intent; verbatim preserves meaning
Include secrets, tokens, or API keys	Security risk
Include full file contents from the workspace	Too large; export should be <10KB
Omit the completeness statement	User needs to know if anything was missed