Security audit

软件测试用例设计

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only test-case writing skill with some imperfect sample credential placeholders, but no evidence of hidden execution, data theft, or unsafe persistence.

Install only if you want a Chinese-language helper for producing structured test cases. Before sharing generated test cases, replace any sample usernames, passwords, bearer tokens, or API keys with placeholders, and be aware that requests mentioning Agent may invoke this skill even when you only intended a general agent discussion.

SkillSpector

By NVIDIA

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
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The example includes realistic-looking account credentials in a shared formatting spec, which can normalize unsafe secret-handling practices and be mistakenly copied into real test artifacts or prompts. Even if the values are illustrative, embedding credential-shaped data exceeds the minimum needed to explain output formatting and increases the chance of accidental secret reuse or disclosure.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The API test example embeds a bearer-token-shaped value directly in the document, which can be copied into downstream materials and encourages unsafe handling of authentication data. A format specification does not need token-like content to teach structure, so including it creates avoidable secret-exposure risk without functional benefit.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The trigger logic is overly broad because terms like "Agent" can appear in many unrelated user requests, and the default fallback activates the skill even when intent is unclear. This can cause the skill to engage outside its intended scope, leading to inappropriate file loading, irrelevant guidance, or unsafe assistance patterns in contexts involving agent security testing.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The manifest description declares a very broad set of applicable scenarios without clear exclusions, boundaries, or examples of when the skill should not activate. In practice, this increases accidental invocation and scope creep, which is risky because the skill may influence outputs in situations not limited to test-case authoring.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: Showing username/password-style values without any safety warning can lead users to treat secrets casually in test documentation and prompts. In a broadly reused reference file, this omission increases the risk that real credentials are later documented, shared, or committed in similar form.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The bearer token example lacks any warning about secret exposure, making it more likely that users will paste actual tokens into specs, tickets, or repositories. Because this is a formatting reference, the omission is especially unnecessary and spreads insecure documentation habits across many generated test cases.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: references/examples/format-spec.md:104