Mimiq Memory

Mimiq Memory

Persistent memory, multi-profile identity, and feedback system for Mimiq. All data is stored locally at ~/.claude/mimiq-memory/ — never sent anywhere.

Database Structure

~/.claude/mimiq-memory/
├── profiles.json          # User identity + platform map
├── fingerprints.json      # Style fingerprint versions + feedback deltas
├── posts.db               # Every post written, rated, published
├── audits.db              # Every audit run + corrections
└── shared_patterns.json   # Opt-in community intelligence (anonymized)

All reads/writes go through the scripts in this skill's scripts/ directory. Always run python3 scripts/init_db.py first to ensure the database exists.

Step 1 — Session Start: Load Memory

At the start of any Mimiq session, run:

python3 ~/.claude/skills/mimiq-memory/scripts/load_memory.py

This returns:

• The user's profile map (all linked platforms + handles)
• The latest style fingerprint
• Summary of past audits
• Any pending feedback requests

If no memory exists yet, initialize with:

python3 ~/.claude/skills/mimiq-memory/scripts/init_db.py

Then prompt the user to set up their profile (see Step 2).

Step 2 — Profile Setup (First Time)

If no profile exists, ask:

"Let's set up your Mimiq profile. What's your name or handle? And which platforms are you on? I'll link them all to one identity."

Collect:

• Display name / preferred handle
• Platform handles (LinkedIn, X, Reddit, Medium, Substack, Instagram, Facebook)

Save with:

python3 ~/.claude/skills/mimiq-memory/scripts/save_profile.py \
  --name "Salman" \
  --linkedin "m-salman-shahid" \
  --twitter "slayerman420" \
  --reddit "slayerman420" \
  --medium "msalman"

Multi-profile linking: If the user mentions a new handle during any session, run:

python3 ~/.claude/skills/mimiq-memory/scripts/link_platform.py \
  --platform "substack" \
  --handle "salman-writes"

Mimiq will automatically treat all linked handles as the same person across all skills.

Step 3 — Auto-Save After Each Skill

After any Mimiq skill completes, save the output automatically:

After voice-capture:

python3 ~/.claude/skills/mimiq-memory/scripts/save_fingerprint.py \
  --data '<fingerprint JSON>'

After content-audit:

python3 ~/.claude/skills/mimiq-memory/scripts/save_audit.py \
  --platform "linkedin" \
  --data '<audit JSON>'

After post-strategist writes a post:

python3 ~/.claude/skills/mimiq-memory/scripts/save_post.py \
  --platform "linkedin" \
  --topic "product launch" \
  --content '<post text>'

Each saved item gets a unique ID that can be used for feedback later.

Step 4 — Feedback System

Post Feedback

After every ghostwritten post, ask:

"How did this feel? (1) Sounds exactly like me (2) Pretty close (3) Off — here's why: ___"

Save feedback with:

python3 ~/.claude/skills/mimiq-memory/scripts/save_feedback.py \
  --type post \
  --id <post_id> \
  --rating <1-3> \
  --notes "too formal in the opening"

Audit Feedback

After every audit, ask:

"Does this audit feel accurate? Anything missing or wrong?"

Save with:

python3 ~/.claude/skills/mimiq-memory/scripts/save_feedback.py \
  --type audit \
  --id <audit_id> \
  --notes "missed my Substack — handle is different"

Fingerprint Feedback

If the user says "that doesn't sound like me" at any point:

python3 ~/.claude/skills/mimiq-memory/scripts/save_feedback.py \
  --type fingerprint \
  --id <fingerprint_id> \
  --notes "I never use rhetorical questions"

Fingerprint feedback automatically updates the active fingerprint rules.

Step 5 — Memory Queries

Users can query their history at any time:

Step 6 — Community Intelligence (Opt-In)

Ask the user once:

"Would you like to contribute anonymized patterns to the Mimiq community? This helps surface insights like 'posts under 100 words get 3x more engagement on LinkedIn' — validated across real users. Your personal data is never shared, only aggregated patterns. (Yes / No / Ask me later)"

If yes, save opted-in status and begin contributing anonymized engagement patterns after audits.

View community insights:

python3 ~/.claude/skills/mimiq-memory/scripts/community_insights.py

This surfaces patterns like:

• "LinkedIn posts published Tuesday 8–10am average 40% more engagement (n=127 users)"
• "Posts with personal stories get 2.8x more comments than tips-only posts (n=89 users)"

Step 7 — Session End: Save State

At the end of every session, run:

python3 ~/.claude/skills/mimiq-memory/scripts/save_session.py

This snapshots the current fingerprint version, logs what was done, and flags any pending feedback items for the next session.

Reference Files

• references/schema.md — Full database schema for all tables
• scripts/init_db.py — Initialize database
• scripts/load_memory.py — Load full memory state
• scripts/save_profile.py — Save/update user profile
• scripts/link_platform.py — Link new platform handle to existing profile
• scripts/save_fingerprint.py — Save new fingerprint version
• scripts/save_audit.py — Save audit result
• scripts/save_post.py — Log a ghostwritten post
• scripts/save_feedback.py — Save feedback on any item
• scripts/save_session.py — Save end-of-session snapshot
• scripts/query_posts.py — Query post history
• scripts/query_audits.py — Query audit history
• scripts/query_feedback.py — Query feedback history
• scripts/query_fingerprint.py — Query current fingerprint
• scripts/query_profile.py — Query user profile
• scripts/community_insights.py — Show community patterns