agent-memory-systems
Design tenant-scoped memory, retrieval and retention for agents
by sickn33Free
Description
agent-memory-systems is an agent skill for designing tenant-scoped memory, retrieval and retention for AI agents. It is for planning retrieval-augmented generation, conversation history, vector stores and memory that persists across sessions. The skill separates three kinds of record: semantic records for facts and preferences, episodic records for events, and procedural records for reusable methods. It keeps retrieval quality distinct from storage volume, and it binds persistent access to a verified identity. Its procedure begins by establishing the task, the authenticated tenant or user, permission to retain content, retention limits and the existing store and model versions. It then specifies record identifiers, source evidence, timestamps, namespaces, retrieval limits and a policy for conflicts and versions. It preserves original records and proposes consolidation without deleting anything. Testing covers rejection across identities, stale or conflicting records, bounded retrieval and failed writes, using isolated test data, and benchmarks retrieval against held-out questions before storage grows.…
Highlights
- Tenant-scoped memory design
- Semantic, episodic and procedural records
- Retrieval and retention policies
- Preserves originals, no deletion
- Cross-identity rejection tests
Ratings & Reviews
No ratings yet
Be the first to rate agent-memory-systems.
People also visit
loop-design-check
Design a goal-oriented agent loop or review one for failure modes: spinning
agent-memory-systems FAQ
What kinds of memory does it cover?
Semantic records for facts and preferences, episodic records for events, and procedural records for reusable methods.
Does it delete old memories?
No. The skill preserves original records and proposes consolidation without deletion.
How is access controlled?
Persistent access is bound to a verified identity, and cross-identity access is tested for rejection.
Are the guide's code sketches deployable?
Not all of them. The skill says to choose an adapter that the installed SDK supports.
How do I install it?
Run npx skills add https://github.com/sickn33/agentic-awesome-skills --skill agent-memory-systems.
Is it checked for security?
Its security audits show passes from Gen Agent Trust Hub, Socket and Snyk.
What are the best agent-memory-systems alternatives?
Popular alternatives to agent-memory-systems include hindsight-docs, eval-harness and iterative-retrieval. See all agent-memory-systems alternatives