concurrencyguard

Security checks across malware telemetry and agentic risk

Overview

ConcurrencyGuard is a local code-scanning skill with disclosed git-hook and baseline features, but users should treat its findings as heuristic and review persistent workflow changes before enabling them.

Install only if you want a local heuristic concurrency scanner. Review hook installation before running it because it changes commit behavior, review baseline files because they suppress known findings, and treat reported race-condition findings as leads to verify rather than definitive vulnerabilities.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (10)

Context-Inappropriate Capability

Medium

Confidence: 82% confidence
Finding: The team report performs `git blame` author attribution for each finding, which goes beyond stated concurrency analysis and can expose contributor identity and performance-like analytics in generated output. In enterprise or shared environments, this can leak personal/workflow metadata and create privacy or misuse risks, especially if reports are broadly distributed or uploaded to CI artifacts.

Intent-Code Divergence

Medium

Confidence: 98% confidence
Finding: The comments describe JWT-based offline validation, but the implementation only performs full HMAC signature verification when CLAWHUB_JWT_SECRET is present. Otherwise, any attacker who can supply a syntactically valid three-part token with a non-empty signature can forge claims such as tier or expiry, enabling license bypass and undermining trust in authorization checks across the skill.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The LK-003 rule claims to detect missing unlock protection, but the regex only matches any Lock()/RLock() call. In a security analyzer, this causes broad false positives that desensitize users to findings and can hide real concurrency defects behind noisy output.

Intent-Code Divergence

Medium

Confidence: 94% confidence
Finding: The LK-006 rule treats all sync.Mutex usage as a problem even though using Mutex is often correct and safe. This undermines analyzer accuracy and can pressure users into inappropriate locking changes, reducing trust in truly important warnings.

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The LK-009 rule says it detects missing lock usage in Python, but it merely matches an import of threading. That creates misleading high-severity findings for benign code and increases alert fatigue in a tool that users may rely on for secure concurrency review.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The AW-002 rule equates any .then()/.catch() usage with a missing await, but promise chaining is often intentional and safe. Misclassifying common async patterns as critical can overwhelm users with bad guidance and obscure actual unhandled-async or fire-and-forget defects.

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The TS-001 rule labels a bare static instance field as a thread-unsafe singleton, but that declaration alone does not prove singleton behavior or unsafe publication. In a concurrency security tool, such overclaiming produces misleading critical alerts and reduces the credibility of genuine race-condition findings.

Intent-Code Divergence

Medium

Confidence: 98% confidence
Finding: The TS-007 rule is internally contradictory: it says shared Pattern/Regex objects are unsafe, while the recommendation correctly notes Pattern is thread-safe and only Matcher is not. This can directly mislead developers into 'fixing' safe code and distract from real thread-safety issues involving mutable matcher state.

Missing User Warnings

Low

Confidence: 77% confidence
Finding: Installing a pre-commit hook changes repository behavior and can block or alter the user's normal commit flow on future commits. Even though this is user-invoked and documented, the lack of a prominent warning/confirmation increases the risk of surprising persistent modification to the developer environment.

Missing User Warnings

Low

Confidence: 79% confidence
Finding: The baseline command creates a suppression file that can affect future scan visibility, and if overwritten silently it may erase prior review decisions or mask issues. In a security-analysis tool, unexpected baseline creation is especially sensitive because it can reduce detection fidelity over time.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal