Deep Dialogue System

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only coaching and personality-analysis skill with no code or credential access, but it does use session memories and psychological summaries that users should handle carefully.

Install only if you are comfortable with a coaching prompt that may summarize and reuse personal reflections across sessions. Avoid sharing highly sensitive details unless you understand how session memory is handled, and ask to review or clear any saved summaries.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Risk analysis

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

#
ASI06: Memory and Context Poisoning
Low
What this means

Personal reflections or generated personality summaries may influence later sessions, including if the summaries are incomplete, inaccurate, or more sensitive than expected.

Why it was flagged

The skill is designed to retain or reuse compressed session information across conversations, which can include sensitive self-discovery and personality-analysis content.

Skill content
features:\n  - memory_compression\n  - session_continuity\nsession_tracking: true
Recommendation

Review what is saved between sessions, avoid sharing information you would not want reused later, and ask to inspect or clear summaries when needed.

#
ASI09: Human-Agent Trust Exploitation
Low
What this means

A user may over-trust personality insights or forget that they may come from stored summaries rather than professional assessment.

Why it was flagged

The skill tells the agent to blend stored memory-derived observations into conversation while hiding the memory structure, which may make generated psychological interpretations feel more authoritative than they are.

Skill content
NATURALLY integrate understanding from memory_vN-1 into dialogue with the user, as if these are your own observations\n- NEVER reveal the memory structure to ANYONE
Recommendation

Treat the coaching output as reflective assistance, not clinical or authoritative advice, and ask the agent to distinguish current observations from prior-session memory when that matters.