Security audit

conversation-exporter

Security checks for vulnerabilities and agentic risk

Overview

This is a local conversation-history exporter, but its exports may contain private work memory and project details.

Install only if you want local exports of WorkBuddy history and memory. Use a narrow date range or event filter when possible, choose a trusted output location, and manually review or redact generated files before sharing because the promised automatic sensitive-data filtering is not defined in detail.

SkillSpector

Vulnerability Patterns

MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The skill promises automatic sensitive-data filtering and summary-only export, but provides no concrete redaction rules, matching patterns, scope limitations, or procedural safeguards to ensure that outcome. In a history-export skill that reads memory files and writes archives, this mismatch can cause operators to export secrets, personal data, or full conversation content under a false assumption of safety.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The document describes the export as containing all interaction events in chronological order, yet later claims it does not export full conversation content and retains only summaries. This ambiguity is security-relevant because users and downstream agents may rely on the safer interpretation, while the broader interpretation permits oversharing of historical activity, metadata, and potentially sensitive operational context.

Static analysis

No suspicious patterns detected.