AgentReadyHomeAgent Listing

← remembering-conversations

remembering-conversations — agentic threat model

8.2AIVSS 8.2 · High

This agent acts as a local read surface over historical Claude Code sessions, presenting a high confidentiality risk if malicious prompt injections or unauthorized tools exploit its semantic search to exfiltrate sensitive past decisions, code, or credentials.

OWASP AIVSS score rationale

AIVSS = (CVSS_Base + AARS) × Mitigation_Factor, where AARS = (10 − CVSS_Base) × (Factor_Sum / 10) × ThM
CVSS base 7.5AARS uplift 0.75Factor sum 3.0/10Threat ×1.0Mitigation ×1.0
Autonomy of Action
0.30
Goal-Driven Planning
0.10
Self-Modification
0.10
Dynamic Tool Use
0.20
Persistent Memory
0.90
Contextual Awareness
0.70
Dynamic Identity
0.00
Multi-Agent Interactions
0.00
Non-Determinism
0.40
Opacity & Reflexivity
0.30

Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.

MAESTRO 7-layer threat model

Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.

L1 · Foundation Models⚠ not certain from listing

Not certain from the listing — assumes Claude (via Claude Code) is the underlying foundation model. Threats include indirect prompt injection where malicious data retrieved from past conversations reprograms the model's current session behavior.

L2 · Data Operations✓ mapped

Directly handles local session data and vector/text search. Primary threats include data exfiltration of sensitive historical conversations (e.g., hardcoded keys or proprietary code stored in history) and local database tampering/poisoning.

L3 · Agent Frameworks✓ mapped

Integrates as a skill within Claude Code. Threats include memory poisoning, where a malicious file or input in a previous session is indexed and later retrieved to exploit the orchestration framework during a search query.

L4 · Deployment & Infrastructure✓ mapped

Runs locally on the user's machine as a Claude Code skill. Threats include unauthorized local file access to the conversation database and privilege escalation if the host environment is not properly sandboxed.

L5 · Evaluation & Observability⚠ not certain from listing

Not certain from the listing — there is no mention of built-in logging, guardrails, or evaluation mechanisms to detect anomalous search queries or unauthorized history harvesting.

L6 · Security & Compliance (cross-cutting)⚠ not certain from listing

Not certain from the listing — as a free, open-source community skill, it lacks explicit access control policies, enterprise compliance certifications, or governance frameworks to restrict history access.

L7 · Agent Ecosystem⚠ not certain from listing

Not certain from the listing — while it operates within the Claude Code ecosystem, it is unclear if other third-party tools or agents can silently invoke this skill to read the user's history without explicit consent.

MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).