Security audit

Soccer Predict

Security checks for vulnerabilities and agentic risk

Overview

The skill does football betting analysis, but it may save prediction history and learned model changes locally without clear opt-in or cleanup controls.

Install only if you are comfortable with a skill that browses titan007.com for match data and may keep football prediction history and learned weights in local memory files. Before using post-match review, require the agent to ask before writing memory, and treat all betting outputs as uncertain analysis rather than financial or gambling advice.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (7)

Description-Behavior Mismatch

Medium

Confidence: 91% confidence
Finding: The framework explicitly instructs the agent to persist learned models, league profiles, and match history into local memory files across sessions. That creates undeclared statefulness and data retention beyond a one-shot prediction/review workflow, which can surprise users, expand the attack surface, and cause unintended accumulation of potentially sensitive or user-linked data.

Context-Inappropriate Capability

Low

Confidence: 80% confidence
Finding: The skill directs the system to retain cross-session league-specific profiles and historical learning memory, introducing long-term behavioral state not clearly necessary for the advertised function. While this appears aimed at model improvement rather than abuse, undocumented retention can still create privacy, transparency, and integrity concerns if users do not expect their interactions to shape persistent future behavior.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The README states that the skill automatically scrapes titan007.com and performs post-match learning and weight optimization, but it does not warn users that invoking the skill may trigger external network access or persistently modify internal state. In an agent setting, undisclosed outbound requests and silent state changes can violate user expectations, create privacy/compliance issues, and make behavior less auditable.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The trigger conditions are broad enough to activate on many ordinary football-related requests, including generic analysis, prediction, handicap, and result-review queries. Over-broad activation can cause unintended scraping, betting advice generation, and persistence behavior to run without sufficiently explicit user intent, increasing privacy, compliance, and safety risk in a gambling-related skill.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill states it saves framework updates and match history to memory for cross-session learning, but the description does not prominently warn users that data persists across sessions. Hidden persistence undermines informed consent and can retain user-supplied content or derived behavioral data longer than expected, which is especially sensitive in a betting-advice context.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: This file explicitly instructs the system to generate betting recommendations, expected value calculations, and confidence levels, but it does not include any warning that gambling can cause financial loss, that outputs may be inaccurate, or that users should not rely on them as financial advice. In the context of a betting-prediction skill, that omission increases the chance that users treat speculative model output as trustworthy decision support and incur monetary harm.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The framework instructs persistent writes to local memory files without any user-facing disclosure or confirmation. Silent modification of local data is a meaningful security and trust issue because it changes system state outside the immediate response, can interfere with other sessions or tools that read the same memory area, and removes informed consent from the user.

Static analysis

No suspicious patterns detected.