Back to skill

Security audit

smart-search-zh

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a legitimate AISA search client, but it exposes broader third-party search, URL extraction, and synthesis features than the skill description clearly discloses.

Install only if you are comfortable sending search queries, URLs, and retrieved result bundles to AISA's external API under your AISA_API_KEY. Avoid using it with secrets, private/internal URLs, regulated data, or prompts that should remain local, and prefer explicit subcommands over broad automatic invocation.

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
  • 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 Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (11)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The declared purpose says this is a narrowly scoped hybrid web+academic search skill, but the backing client reportedly supports broader modes such as other search providers, URL extraction, and multi-source analysis. This mismatch is dangerous because operators and users may approve or invoke the skill under false assumptions, enabling unreviewed data collection or broader network behavior than the manifest suggests.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill declares access to an environment variable and implies networked execution via a Python client, but it does not explicitly scope or constrain which tools/actions are permitted. That omission weakens reviewability and increases the chance the skill is invoked with broader runtime capabilities than users expect, especially since it can reach external services using a secret.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The activation condition is phrased broadly around any need to balance public-web coverage and academic depth, which leaves substantial room for subjective interpretation. Ambiguous triggering can cause the skill to run in contexts where external search or secret-backed API use was unnecessary, increasing accidental data exposure and unintended third-party queries.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

Saying the skill is a 'good default choice' when questions involve both news and papers encourages automatic selection without verifying that external querying is appropriate. In skill-routing systems, vague default guidance can expand the blast radius of a networked skill and make unintended data transfer to third-party services more likely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest sets the skill language to "zh" with no accompanying indication that the user can choose another language or that the skill is region-specific. This is a natural-language policy concern because it imposes a locale/language constraint without visible opt-in or documented justification in the file.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The file advertises a smart hybrid search client but also exposes additional capabilities including generic web search, scholar search, Tavily search, raw URL extraction, Sonar research, and synthesis flows. In an agent-skill context, this scope expansion increases the attack surface and enables behaviors not clearly covered by the stated skill purpose, which can lead to unintended data handling or policy bypass through hidden or loosely documented commands.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
89% confidence
Finding

The client is explicitly designed to transmit user queries, URLs, and aggregated search results to an external API service. External transmission is expected for a search client, but in an agent-skill environment it remains security-relevant because potentially sensitive prompts, targets, and retrieved content leave the local trust boundary and may be processed by third parties.

Content

Scanner excerpt · scripts/search_client.py (reported line 28)May include surrounding context.

python
from concurrent.futures import ThreadPoolExecutor, as_completed
from typing import Any

AISA_BASE = "https://api.aisa.one/apis/v1"


def get_api_key() -> str:

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The extract mode accepts arbitrary user-supplied URLs and sends them to an external extraction service, then prints raw content locally. In an agent setting, this can be abused to exfiltrate sensitive internal URLs or retrieve and disclose content from resources the user did not intend to expose, especially if the surrounding system has access to private links or tokens.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The Sonar/Perplexity mode introduces a separate research-model querying capability that is not part of the stated smart web-plus-scholar search function. This expands external transmission paths and may send user prompts or sensitive research questions to another backend without clear disclosure, creating data-governance and least-privilege concerns.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This code explicitly sets the synthesis language to "en" when calling the explain endpoint, which forces a specific locale regardless of user preference. The file does not offer a language option or document a justified region-specific constraint, so it violates the language/locale policy criteria.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
76% confidence
Finding

The function documentation describes the command as a search-and-confidence workflow. In addition to that, the code sends aggregated results to /scholar/explain and generates a synthesized summary with citations, which is a materially different behavior than the docstring states.

Content

No source excerpt is available for this finding.