pa-memory-delegation
Skill8 starsChanged 43 days agoMITView on GitHub
Use when the user tells you something worth remembering about their life (facts, events, people, preferences), or asks what you know about them. Manages the USER's memory (life events, facts, learnings) - NOT the AI's operational memory. Uses a hybrid Memvid + SQLite + Markdown architecture for fast semantic search with mutable metadata tracking.
- Installs packages
In this repository4 other files it gives its agents.
- brain/agents.mdAGENTS.md
- file-processingSkill
- mail-handlerSkill
- pa-onboardingSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 40
The user corrects you: use correct with the right statement.
- Very specific: line 41
Unsure or passing remarks: add them with --short-term.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
PA Memory Delegation
This is the user's memory, kept on their behalf: their life, health, family, money, decisions, preferences. It is not your own operational memory (CLAUDE.md, brain/*.md), which you manage separately.
It has three layers:
- Memvid indexes (
indexes/memories.mv2long-term,indexes/short-term.mv2staging): semantic search. Append-only. - SQLite (
indexes/memory_meta.db): the mutable part — confidence, how often a memory was reinforced and retrieved, supersessions, soft deletes. - Markdown in
memory/: the human-readable record (profile,journal-YYYY.md,goals.md).
Commands
memory.sh (in $LOLABOT_HOME/tools/, or ~/.local/bin if installed there) wraps tools/memory_indexer.py and activates the virtualenv.
memory.sh add "<content>" --type <type> [--date YYYY-MM-DD] [--people "A,B"] [--tags "x,y"] \
[--source "<where it came from>"] [--files "<paths>"] [--confidence 0.0-1.0] [--context "<why>"] \
[--topic-key <key>] [--short-term] [--force]
memory.sh find "<query>" [--type <type>] [--year YYYY] [--person "<name>"] [--limit 10] [--short-term]
memory.sh correct "<corrected statement>" [--topics "a,b"] [--source "<who said so>"]
memory.sh remove <frame-id> # soft delete
memory.sh review | promote [--dry-run] # short-term staging
memory.sh stale [--refresh] | statsTypes: fact, event, person, preference, goal, decision, learning, note (default), pattern, insight.
Saving
Save what the user would expect you to remember next month: facts about them and their family, dated events, people and how they relate to the user, preferences, goals, decisions and why. Skip conversational filler. Credentials (passwords, PINs, keys) never go into memory.
- Dated things get
--date, and the people involved get--people, sofind --yearandfind --personwork. - Facts that change (address, employer, a balance): give them a stable
--topic-keysuch ashome-address. A new value under the same key supersedes the old one instead of sitting beside it. - The user corrects you: use
correctwith the right statement. It is stored at high confidence and outranks the memories it contradicts. - Unsure or passing remarks: add them with
--short-term. Short-term memories move to long-term withpromoteonce reinforced twice or at 0.9 confidence.
Saving the same fact again reinforces it (confidence +0.05, from 0.7) instead of duplicating it. The duplicate check compares the beginning of the text, so a reworded fact is stored as a new memory: when you know a fact is already there, phrase it the same way or use a topic key.
Also write the information to the matching markdown file, as CLAUDE.md requires: personal facts to the user's profile, life events to memory/journal-YYYY.md, goals to memory/goals.md.
Recalling
When the user asks what you know, or a task depends on their history, run find before answering, and say where a fact came from when it matters. Results show confidence and how often each memory was reinforced; low-confidence or stale memories (memory.sh stale) are worth confirming with the user rather than stating as fact.
Troubleshooting
memvid-sdk not installed:source .venv/bin/activate && uv pip install memvid-sdk.- Index not found: the first
addcreates it. - SQLite metadata empty for old memories:
memory.sh migrate.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

