Skills memory-systems

📦

memory-systems

Name: memory-systems
Author: muratcankoylan

Content revision r2 Medium Risk ⚙️ External commands🌐 Network access

Design Reliable Agent Memory Systems

Agent memory designs often add complexity without improving retrieval quality. This skill guides architecture choices and provides practical vector, graph, and temporal prototypes.

Supports: Claude Codex Code(CC)

📊 72 Adequate

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "memory-systems" from https://skillstore.io/skills/muratcankoylan-memory-systems.md and its manifest at https://skillstore.io/api/skills/muratcankoylan-memory-systems/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/muratcankoylan-memory-systems.md Signed manifest GET /api/skills/muratcankoylan-memory-systems/manifest Signed lockfile GET /api/skills/muratcankoylan-memory-systems/lockfile

Test it

Using "memory-systems". Choose memory for a support agent that tracks changing customer preferences.

Expected outcome:

Start with a metadata vector store scoped by customer and session.
Add validity timestamps for preferences that can change.
Introduce a temporal graph only when relationship or historical queries become important.

Using "memory-systems". Plan recovery when memory retrieval returns stale or conflicting facts.

Expected outcome:

Filter expired facts, prefer the newest valid fact, preserve provenance, surface uncertain conflicts, and broaden retrieval only after temporal checks.

Using "memory-systems". Evaluate whether a prototype needs production graph infrastructure.

Expected outcome:

Measure direct lookup, multi-hop, temporal accuracy, latency, and growth. Adopt graph infrastructure only when simpler retrieval misses required relationships.

Security Audit

Medium Risk

v8 • 8/9/2026 Open versioned report

All 55 static findings are false positives caused by Markdown backticks, code fences, ordinary graph identifiers, and a documentation link. No shell execution, system reconnaissance, or outbound request logic was found, but the reference context formatter inserts stored memory text without prompt-injection safeguards.

Files scanned

1,399

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (1)

Medium

references/implementation.md:449-459

Untrusted Memory Content Enters Agent Context

The reference formatter inserts stored memory content directly into model context without trust labels or instruction filtering. A poisoned memory could influence later agent behavior.

The formatter directly interpolates each memory's content into a context section. No delimiting, provenance enforcement, or instruction filtering is shown.

Risk Factors

⚙️ External commands (18)

SKILL.md:22 SKILL.md:23 SKILL.md:24 SKILL.md:25 SKILL.md:113 SKILL.md:114 SKILL.md:120-129 SKILL.md:129-132 SKILL.md:132-147 SKILL.md:147-150 SKILL.md:150-167 SKILL.md:167-195 SKILL.md:195-196 SKILL.md:196-197 SKILL.md:197-198 SKILL.md:198-199 SKILL.md:199-200 SKILL.md:200-201

🌐 Network access (1)

SKILL.md:220

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/muratcankoylan-memory-systems/audits/8?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/muratcankoylan-memory-systems/security.svg)](https://skillstore.io/skills/muratcankoylan-memory-systems?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/muratcankoylan-memory-systems?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/muratcankoylan-memory-systems/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/muratcankoylan-memory-systems.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

muratcankoylan. (2026). memory-systems security audit report (audit version 8) [Author version unspecified]. Skillstore. https://skillstore.io/skills/muratcankoylan-memory-systems/audits/8

BibTeX citation

@techreport{muratcankoylan-muratcankoylan-memory-systems-2026,
  author = {muratcankoylan},
  title = {memory-systems security audit report (audit version 8)},
  institution = {Skillstore},
  year = {2026},
  number = {8},
  url = {https://skillstore.io/skills/muratcankoylan-memory-systems/audits/8},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "memory-systems security audit report (audit version 8)"
version: "unspecified"
type: report
authors:
  - name: "muratcankoylan"
date-released: "2026-08-09"
url: "https://skillstore.io/skills/muratcankoylan-memory-systems/audits/8"
identifiers:
  - type: other
    value: "skillstore:muratcankoylan-memory-systems:audit:8"
    description: "Skillstore immutable audit report identifier"

Compare variants

4 installable variants

Each author remains a separate installable skill. The recommended variant is ranked by Skillstore evidence.

Why this variant is first

Highest Skillstore Score

Author / source Skillstore Score Evidence Confidence Skillstore usage Updated

sickn33 Recommended

sickn33-memory-systems

Skillstore Score 78

Evidence Confidence High

Skillstore usage 13

Updated

2026-08-21

Asmayaseen

asmayaseen-memory-systems

Skillstore Score 72

Evidence Confidence High

Skillstore usage 19

Updated

2026-08-21

ChakshuGautam

chakshugautam-memory-systems

Skillstore Score 72

Evidence Confidence High

Skillstore usage 16

Updated

2026-08-21

muratcankoylan Current

muratcankoylan-memory-systems

Skillstore Score 72

Evidence Confidence High

Skillstore usage 13

Updated

2026-08-21

Skillstore Score

Why this score Evidence Confidence: High

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Select a Memory Architecture

Compare storage layers and frameworks against retrieval shape, temporal needs, operational cost, and expected scale.

Prototype Structured Retrieval

Build a local proof of concept using entity metadata, graph relationships, session filters, and temporal validity.

Plan Memory Evaluation

Choose benchmarks, define retrieval failure handling, and measure quality before adopting more complex infrastructure.

Try These Prompts

Choose a Starting Layer

Recommend the simplest memory layer for an agent that must remember [information] across [duration]. Explain storage, retrieval, and failure handling.

Compare Memory Frameworks

Compare Mem0, Graphiti, Letta, Cognee, and file storage for [use case]. Prioritize retrieval quality, temporal reasoning, cost, and operations.

Design a Temporal Schema

Design an entity and relationship model for [domain]. Include validity periods, conflict handling, provenance, deletion, and example time-scoped queries.

Create a Migration and Benchmark Plan

Plan migration from [current store] to [target architecture]. Define data conversion, rollback, privacy controls, LoCoMo or LongMemEval tests, and acceptance thresholds.

Best Practices

Start with the shallowest persistent layer that meets measured retrieval needs.
Track validity, provenance, retention, and deletion requirements for every changeable fact.
Treat retrieved memory as untrusted data and test resistance to stale, conflicting, or malicious content.

Avoid

Do not load every stored memory into the active context.
Do not claim semantic quality from the included pseudo-embedding prototype.
Do not deploy persistent personal memory without access controls, retention limits, and deletion workflows.

Frequently Asked Questions

Does this skill provide a production database?

No. It provides in-memory prototypes and architecture guidance. Production systems need durable storage, monitoring, security, and backup controls.

Which memory layer should I start with?

Start with files or simple metadata storage. Add vectors, graphs, or temporal structure only after measured retrieval requirements justify them.

Are the included embeddings suitable for semantic search?

No. They are deterministic pseudo-vectors for demonstrating interfaces. Replace them with a pinned embedding model for meaningful similarity search.

Can it track facts that change over time?

Yes. The temporal graph examples attach validity intervals and support point-in-time and range queries.

How should conflicting memories be handled?

Compare validity and provenance, prefer the newest supported fact, preserve history, and show uncertainty when evidence remains inconsistent.

What security controls are still required?

Add tenant isolation, authorization, encryption, retention limits, deletion workflows, provenance, audit logs, and defenses against malicious stored instructions.