Agency memory

Skill sethshoultes/great-minds-plugin/skills/agency-memory

Query and manage the Great Minds agency vector store — semantic search over memories stored in SQLite with embeddings. Subcommands - search, add, list, import, export, stats. Filters by type (learning, decision, board-review), agent, project. Use to recall prior decisions, learnings, or critiques across multiple agency projects.From its SKILL.md

Install

npx -y skills add sethshoultes/great-minds-plugin --skill agency-memory

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.6 KB, as published. Nobody here has run it

/agency-memory

Query and manage the Great Minds agency vector store. Memories are stored in SQLite with embeddings for semantic search.

Usage

/agency-memory search "query text"
/agency-memory add --type learning "content here"
/agency-memory add --type decision --agent "Steve Jobs" --project "Dash" "content"
/agency-memory list --type board-review
/agency-memory list --agent "Jensen Huang"
/agency-memory import
/agency-memory export
/agency-memory stats

Instructions

When this skill is invoked, run the memory CLI from the memory-store directory.

Search

Find relevant memories by semantic similarity:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory search "$QUERY" --limit ${LIMIT:-5}

Add

Store a new memory. Valid types: learning, decision, qa-finding, board-review, retrospective, architecture.

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory add \
  --type "$TYPE" \
  --agent "$AGENT" \
  --project "$PROJECT" \
  --content "$CONTENT"

List

List memories filtered by type, agent, or project:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory list --type "$TYPE"
cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory list --agent "$AGENT"
cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory list --project "$PROJECT"

Import

Import existing markdown memories from the Great Minds project:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory import

This parses:

MEMORY.md index (key learnings)
memory/*.md files (operational learnings, architecture decisions, research)
rounds/*/decisions.md (locked project decisions)
rounds/*/board-review-*.md (Jensen Huang board reviews)
rounds/*/qa-report-*.md (Margaret Hamilton QA findings)

Export

Export all memories back to markdown for compatibility:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory export

Stats

Show memory store statistics:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory stats

Maintenance

Prune

Remove duplicate memories (cosine similarity > 0.92) and low-value entries (e.g., "all green" board reviews, "PASS" QA findings):

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory prune
cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory prune --threshold 0.85

Consolidate

Find clusters of similar memories (similarity > 0.75) and merge them into single stronger entries:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory consolidate
cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory consolidate --threshold 0.80

Optimize

Score all memories by recency, uniqueness, and richness. Remove the bottom 10% by score:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory optimize
cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory optimize --percentile 20

Maintain

Run the full maintenance cycle (prune + consolidate + optimize) in sequence. This is what the dream cron calls automatically:

cd "${PIPELINE_REPO:-$(git rev-parse --show-toplevel)}/memory-store" && ./bin/memory maintain

Environment

OPENAI_API_KEY — enables neural embeddings (text-embedding-3-small). Falls back to TF-IDF when absent.
MEMORY_DB — custom path to SQLite database. Defaults to memory-store/memory.db.

Bundled Store

The memory store also ships with this plugin at memory-store/. Any installation gets a fresh copy. Install dependencies before first use:

cd "/path/to/great-minds-plugin/memory-store" && npm install

Design

SQLite with WAL mode for concurrent reads
Cosine similarity search over embedding vectors
Content-hash embedding cache to avoid redundant API calls
Works offline via TF-IDF fallback
Runs locally and on DigitalOcean with no external services