Back to skill

Security audit

tradealpha实时新闻

Security checks for vulnerabilities and agentic risk

Overview

The skill is purpose-aligned for TradeAlpha news lookup, but it asks users to paste an API token into chat and automatically reuses it during the session.

Review before installing if you are not comfortable pasting a TradeAlpha API token into chat. Use a scoped or short-lived token if possible, avoid sharing account-wide credentials, and be aware the skill may trigger on broad news requests involving sources like Reuters or Bloomberg.

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
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (9)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill description includes broad activation phrases such as generic news terms, wire services, and topic-based retrieval requests, which can cause the skill to trigger in contexts beyond explicit user intent. Over-broad routing can expose user prompts or tokens to an unnecessary external news workflow and increase the chance of inappropriate tool invocation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The examples explicitly instruct users to paste a live access token directly into chat, which exposes credentials to conversation history, logging, screenshots, and any downstream systems that may retain prompts. Although the skill says it only uses the token in the current session, that does not mitigate the risk of credential disclosure through the chat channel itself.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

These examples instruct the assistant to solicit a sensitive TradeAlphaToken directly in the conversation before performing API calls. This creates a direct secret-handling anti-pattern: users are trained to disclose credentials in plain chat, increasing the chance of accidental exposure, reuse by unauthorized parties, or capture by logging and telemetry systems.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The later examples reinforce retaining and reusing the user's token across the session, which increases the blast radius if the conversation is exposed and encourages persistence of a sensitive secret in model context. In this skill context, the danger is heightened because the behavior is presented as the normal operating flow for an authentication credential, not an incidental example.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly instructs users to paste an API token into chat and gives operational guidance for using that secret, but provides no minimization, masking, or safer authentication flow. Secrets provided in natural-language chat can be exposed in conversation history, logs, analytics, or to downstream components, creating avoidable credential-handling risk.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The instruction to remember and automatically reuse a user-provided token across the session creates secret persistence in conversational state. Persisting credentials in natural-language context increases the chance of accidental disclosure, prompt leakage, cross-turn misuse, or unintended reuse in later requests beyond the user's immediate intent.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Repeatedly directing the model to ask for tokens in chat normalizes secret sharing in an unsafe channel and increases the likelihood that sensitive credentials will be exposed in transcripts or mishandled by the system. The danger is amplified because the skill operationally depends on this pattern before every API call when no cached token exists.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instruction “用中文总结重点内容” forces a specific language for responses and does not offer the user a choice. This is a natural-language policy issue because it imposes a locale/language preference without opt-in or documented justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The failure message is hard-coded in Chinese (请直接在聊天里发送 TradeAlphaToken...) and the document repeatedly uses Chinese-only example phrases and category values without stating that responses can be localized based on user preference. This can violate language/locale policy when the skill is used in a broader multilingual environment.

Content

No source excerpt is available for this finding.