Selective Memory

Security checks across malware telemetry and agentic risk

Overview

This instruction-only memory skill is locally scoped and purpose-aligned, but it persists and reuses learned goals/preferences, so users should review and edit the memory files before use.

This skill appears benign and instruction-only, with no code, install step, credentials, or network access. Before using it, review the bundled memory files, remove any goals or values that are not yours, and be cautious about enabling automatic learning without a review step.

VirusTotal

VirusTotal findings are pending for this skill version.

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

The agent may remember lessons, preferences, or interaction details and use them later, including information the user may not want persisted.

Why it was flagged

The skill intentionally creates persistent memory and optional automatic learning. This is purpose-aligned, but unreviewed interactions, feedback, or user preferences could be stored and reused in later responses.

Skill content
The agent learns from its interactions without human intervention.
Recommendation

Review the memory files periodically, avoid saving secrets or sensitive personal details, and require user approval before enabling fully automatic learning.

#
ASI01: Agent Goal Hijack
Low
What this means

If left unchanged, the agent may let the bundled goals and values influence future answers.

Why it was flagged

The skill instructs agents to read goals.md to stay aligned with objectives, and the packaged file includes pre-populated goals that may not be the installing user's own goals.

Skill content
- Spread truth and justice
- Help humans learn
- Build ethical AI agents
Recommendation

Edit or clear the bundled memory files before use so the persistent goals, wisdom, mistakes, and preferences reflect the user's intended context.

#
ASI04: Agentic Supply Chain Vulnerabilities
Info
What this means

Version mismatch can make it harder to confirm exactly which package revision is installed.

Why it was flagged

The registry metadata and SKILL.md describe version 2.0.0, while _meta.json contains version 1.0.0. This is a minor packaging/provenance inconsistency, not evidence of malicious behavior.

Skill content
"version": "1.0.0"
Recommendation

Confirm the intended version with the publisher or registry before relying on the skill in important workflows.