Security audit

OEE AI Cost Tracker

Security checks for vulnerabilities and agentic risk

Overview

This is a local AI usage cost tracker that writes and reads a local JSONL usage log; its privacy risks are real but consistent with its stated purpose.

Install only if you are comfortable keeping a local usage.jsonl file with model names, token counts, costs, task labels, sources, and any descriptions you provide. Avoid putting prompts, secrets, customer data, or proprietary details in the description field, and review or delete the local log when it is no longer needed.

A.I.G

Vulnerability Patterns

Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill describes logging every API call to JSONL but does not prominently warn that local files will contain API usage metadata such as model names, token counts, task labels, and timing context. Even without API keys, such metadata can reveal sensitive business activity, workload patterns, or user behavior if stored insecurely or shared inadvertently.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The trigger phrase "how much have I spent on AI" is broad enough to match normal budgeting or usage questions and could invoke this skill unexpectedly. Over-broad activation increases the chance that the skill runs in contexts where the user did not intend local spend analysis, which can expose or process usage metadata unnecessarily.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The library persists the caller-provided `description` field directly to a local JSONL log file without sanitization, minimization, or any notice that arbitrary text will be stored on disk. Because descriptions may contain prompts, secrets, personal data, or proprietary task details, this creates a real privacy and data-exposure risk if the host, backups, logs, or repository are later accessed by unauthorized parties.

Vague Triggers

Low

Confidence: 88% confidence
Finding: The phrase "track this API call" is vague and could apply to many unrelated debugging or observability tasks. A vague trigger can cause accidental invocation, resulting in unintended logging behavior or confusion about what data is being recorded.

Static analysis

No suspicious patterns detected.