Plur Claw Publish

Security checks across static analysis, malware telemetry, and agentic risk

Overview

This is a coherent persistent-memory plugin, but it can run setup during install, persistently enable itself, and configure an unpinned MCP server outside the reviewed package.

Before installing, review the setup behavior, consider installing with npm lifecycle scripts disabled, inspect any changes to ~/.openclaw/openclaw.json, pin or remove the @plur-ai/mcp npx server, and decide whether auto-learning and auto-capture are appropriate for your conversations.

Static analysis

No static analysis findings were reported for this release.

VirusTotal

VirusTotal engine telemetry is currently stale for this artifact.

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.

#
ASI01: Agent Goal Hijack
Low
What this means

The agent may recall, learn, and capture memories without asking each time.

Why it was flagged

The skill's memory prompt changes the agent's default behavior toward proactive memory-tool use. This is purpose-aligned, but it affects user control over when memory is used.

Skill content
Do not ask permission to use these tools — they are your memory system.\n\nUse these tools proactively, not just when asked:
Recommendation

Install only if you want proactive memory behavior, and disable or edit memory settings if you prefer per-use approval.

#
ASI05: Unexpected Code Execution
Medium
What this means

Installing the package can execute code before the user explicitly enables or reviews the plugin configuration.

Why it was flagged

Installers that honor npm lifecycle scripts can run this setup code automatically during installation, with stderr suppressed and failures ignored.

Skill content
"postinstall": "node dist/setup.js 2>/dev/null || true"
Recommendation

Install with lifecycle scripts disabled if possible, inspect the setup code, and run setup manually only after reviewing the intended config changes.

#
ASI10: Rogue Agents
Medium
What this means

If an OpenClaw config already exists, the plugin may become persistently enabled across future sessions.

Why it was flagged

The setup code can write the OpenClaw config to enable this plugin and set it as the memory slot, creating persistent agent behavior.

Skill content
entries[PLUGIN_ID] = nextEntry; ... slots.memory = PLUGIN_ID; ... writeFileSync(path, JSON.stringify(cfg, null, 2) + "\n", "utf8");
Recommendation

After installation, inspect ~/.openclaw/openclaw.json or OPENCLAW_HOME/openclaw.json and remove the plur-claw entry if it was not intentionally enabled.

#
ASI04: Agentic Supply Chain Vulnerabilities
High
What this means

OpenClaw may later start an unreviewed or changed version of @plur-ai/mcp as part of the agent environment.

Why it was flagged

Setup adds an MCP server that can later run an unpinned npm package via npx -y, meaning code outside the reviewed artifact set may be downloaded or executed.

Skill content
servers.plur = { command: "npx", args: ["-y", "@plur-ai/mcp"], env: { PLUR_PATH: plurPath } };
Recommendation

Require explicit approval before adding this MCP server, pin the exact @plur-ai/mcp version, and review that package separately.

#
ASI06: Memory and Context Poisoning
Medium
What this means

Private, stale, or incorrect conversation details may persist and influence future sessions.

Why it was flagged

The plugin defaults to automatically creating persistent memories and episodic summaries from conversations, which is central to its purpose but sensitive.

Skill content
"auto_learn": { "type": "boolean", "default": true, "description": "Auto-extract learnings from conversations" }, ... "auto_capture": { "type": "boolean", "default": true, "description": "Auto-capture episodic summaries" }
Recommendation

Review where PLUR stores data, disable auto_learn or auto_capture if unwanted, and periodically inspect or delete stored memories.