Back to skill

Security audit

去哪不挤

Security checks for vulnerabilities and agentic risk

Overview

This is a Chinese holiday travel report skill that writes local Markdown and HTML reports as disclosed, with no evidence of exfiltration, persistence, or destructive behavior beyond normal file output.

Install only if you want a Chinese-language, opinionated holiday travel decision report. Expect it to create local Markdown and HTML files in the working directory or specified output directory, including fixed latest filenames that may replace previous generated reports.

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
  • 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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (16)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill requires local file read/write behavior and references local scripts/resources, but it does not declare any explicit tool scope or permission boundaries. That creates an authorization ambiguity where an agent may perform filesystem actions more broadly than intended, increasing the risk of unintended file access or writes in the workspace.

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 the agent to generate multiple local Markdown and HTML files, including a fixed alias file, without informing the user or constraining where writes occur. Unannounced filesystem writes can overwrite existing files, leave persistent artifacts, or be abused to plant misleading content in the workspace.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill enables implicit invocation without any visible activation constraints, so the platform may trigger it on loosely related travel queries without clear user intent. Because this skill produces decision-driving recommendations and writes a local HTML output, ambiguous auto-triggering increases the chance of unintended execution, confusing outputs, or unnecessary local artifact creation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The alias list includes very broad terms such as “周边”, “短途”, “人少”, and “不想太挤”, which can match many ordinary travel conversations unrelated to this specific decision skill. That creates overbroad activation risk: the skill may trigger unexpectedly, capture requests meant for other travel tools, and steer users into its fixed recommendation frames without clear user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Aliases like “海边”, “看海”, “海岛”, and “沙滩” are generic travel concepts that appear in a wide range of benign requests. Using them as direct triggers can cause this skill to intercept general seaside trip planning queries and force a narrow ‘avoid Sanya / prefer Okinawa’ framing even when the user did not ask for anti-crowding or substitution advice.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The activation aliases include broad phrases like “杭州”, “周边不挤”, and “短途更舒服”, which are too ambiguous to safely distinguish this skill from ordinary local travel or city-information requests. This can lead to unintended invocation and biased recommendation behavior, especially because the frame is opinionated and prescriptive rather than informational.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Aliases such as “北京”, “城市”, “大城市”, “citywalk”, “看展”, and “都市” are extremely generic and likely to overlap with many unrelated conversations about urban travel, exhibitions, or lifestyle preferences. In this skill, broad matching is more dangerous because it can redirect general city-trip requests into a precomputed recommendation favoring Seoul or Taiyuan, reducing user control and accuracy.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Terms like “美食”, “舒服”, “慢节奏”, and “放松城市” describe broad preferences rather than a clear request for this skill’s niche function. They are likely to produce accidental activation on general lifestyle or travel chats, after which the skill may assert narrow recommendations that do not reflect the user’s actual constraints or intent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The template title and required output language are entirely in Chinese, and L03 states every output must follow this format, effectively forcing a specific language/locale. There is no user opt-in, alternative language option, or documented region-specific justification for the constraint.

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 generates the report content entirely in Chinese, including the title, section headings, labels, and recommendation text. The script exposes no option for users to select another language or opt in to a Chinese-only locale, which matches the policy concern about forcing a specific language without user choice.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The script accepts a caller-controlled '--output-dir' and an arbitrary '--json-out' path, then writes multiple files without validating that the destinations stay within an approved workspace. In environments where untrusted users can influence arguments, this enables filesystem clobbering of attacker-chosen paths, potentially overwriting application data, user files, or other sensitive artifacts the process can access.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The manifest frames the skill as a decision/reporting tool for holiday travel choices, while this implementation materially acts as a file generator that creates directories and writes multiple Markdown, HTML, and optional JSON files to caller-specified paths. Persisting and overwriting several artifacts is beyond the user-facing description of simply outputting a report, and is not clearly disclosed in the manifest text.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The README explicitly states that the skill will create Markdown and standalone HTML files in the working directory, but it does not warn users about local file creation, possible overwrites, or where artifacts will be written. In an agent context, silent filesystem writes can surprise users, leak sensitive content into local files, or overwrite existing documents if output paths are reused or inferred unsafely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

The document declares lang="zh-CN", and the visible UI text throughout the file is Chinese-only. For SQP-3, this is a natural-language locale policy concern because the file imposes a specific language/locale without offering opt-in or explaining that the skill is intentionally region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown content is entirely written as a Chinese example report and presents the skill output in Chinese without any indication that language is user-selectable. Under the policy rule for natural-language violations, forcing a specific language without opt-in can be a locale-policy issue unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The heading and all example content are written exclusively in Chinese, indicating a fixed language/locale for the skill output. There is no accompanying note that the language is user-selectable or justified as a region-specific requirement, which matches the policy concern for forced locale without opt-in.

Content

No source excerpt is available for this finding.