Security audit

game-negotiation

Security checks for vulnerabilities and agentic risk

Overview

The negotiation engine itself is straightforward, but the package also includes an under-scoped local learning component that can persist notes, preferences, and usage history beyond the core negotiation purpose.

Review this skill before installing if you do not want durable local learning files. The calculation engine appears benign, but the bundled learner should be treated as optional and only used with non-sensitive notes/preferences and a skill directory you intend to modify.

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
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

Findings (4)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 88% confidence
Finding: The skill advertises no permissions, yet the description and analysis indicate file-write capability for local persistence. Undeclared stateful behavior breaks the principle of least privilege and can surprise users or hosting agents, especially if the skill writes usage history or learned patterns without explicit consent.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 96% confidence
Finding: This is a significant description-behavior mismatch: a game-theory negotiation engine is also described as recording usage, errors, user preferences, and persisting learned patterns. Hidden telemetry and preference storage materially expand the data-handling surface, creating privacy, integrity, and trust risks because users invoking negotiation logic would not reasonably expect profiling or local memory persistence.

Description-Behavior Mismatch

High

Confidence: 97% confidence
Finding: The file implements a generic cross-skill persistence and self-learning component that is not necessary for a game-theoretic negotiation engine. Because it can be invoked on arbitrary skill directories and stores operational history and preferences under ~/.workbuddy/skills, it expands the skill’s scope into long-term data collection and modification of other skill state, increasing privacy and supply-chain risk if abused or composed with other tools.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: These functions persist user preferences, error notes, and recent operation history without any visible consent, retention policy, or demonstrated need for the stated negotiation purpose. Free-form note and preference storage can capture sensitive user or workflow data, creating unnecessary privacy exposure and durable local tracking across sessions.

Static analysis

No suspicious patterns detected.