Back to skill

Security audit

Meridian

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent AI intelligence reporter, but it includes unsafe shell-based search examples using user-derived keywords and saves reports by default without clear user control.

Review before installing. Use this skill only in an environment where shell commands are sandboxed or require approval, and prefer built-in web/API tools over copying the curl/python snippets. Confirm report file destinations before use if researched entities, queries, or gathered links should not be retained locally.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:226
Finding

User-Controlled Search Keyword Allows Shell Command Injection

Content
View full analysis
{YYYY-MM-DD}&sort=stars&order=desc&per_page=10" | \ python3 -c " import sys, json data = json.load(sys.stdin) for r in data.get('items', [])[:10]: stars = r['stargazers_count'] name = r['full_name'] desc = (r.get('description') or '')[:100] url = r['html_url'] created = r['created_at'][:10] print(f'{created} | {stars} stars | {name}') print(f' {desc}') print(f' {url}') print() " ``` ### Technical Analysis The skill instructs the agent to extract entity or platform names from the user's request and insert the resulting keyword into shell commands. The `{keyword}` placeholder appears inside a double-quoted shell argument without any required validation, URL encoding, or argument-safe parameter handling. Double quotes do not prevent shell command substitution. If an implementation directly replaces `{keyword}` with attacker-controlled input before running the documented command, constructs such as `$(command)` or backticks are evaluated by the shell. The command substitution executes before `curl` receives the resulting URL. The Hacker News command at lines 201-219 uses the same unsafe interpolation pattern and is vulnerable under the same execution model. ### Attack Path 1. An attacker submits an entity or platform name containing shell metacharacters, such as a command substitution expression. 2. The skill extracts that attacker-controlled value as the search keyword, as directed at lines 105 or 148. 3. The agent directly substitutes the value for `{keyword}` in the shell command. 4. The agent invokes the command through a shell. 5. The shell eval ...[truncated 1442 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (11)

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill embeds shell commands using curl piped into python3, which introduces an execution pathway beyond simple web research. Even if shown as examples, such patterns normalize command execution against dynamic external input and can enable unsafe tool use, broader system interaction, or accidental execution in environments where the agent has shell access.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

This section again directs shell-based API access via curl and local python3 processing, expanding the skill from report generation into arbitrary command execution. The risk is elevated because repository search terms may derive from user input, and shell patterns can be copied or operationalized by an agent with filesystem and network capabilities.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The core operating instructions switch into Chinese and define the assistant persona and behavior in that language, without indicating that language is optional or user-selectable. This can violate language/locale policy because it implicitly constrains interaction to a specific language absent user opt-in.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 224)May include surrounding context.

bash
# 搜索近期创建的高星项目
curl -s "https://api.github.com/search/repositories?q={keyword}+created:>{YYYY-MM-DD}&sort=stars&order=desc&per_page=10" | \
  python3 -c "
import sys, json
data = json.load(sys.stdin)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The skill instructs saving intelligence reports to disk without notifying the user or obtaining consent. This creates an undisclosed side effect and may persist potentially sensitive research targets, queries, or gathered material in a location the user did not expect.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly instructs writing generated intelligence reports into local intel/ files. Even though the content is consistent with the skill’s reporting purpose, persisting outputs to disk without explicit user consent or transparency can create unintended data retention and side effects, especially if prompts or fetched content contain sensitive information.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The file is entirely written as a reporting template in Chinese, including headings, instructions, and example output structure. There is no indication that the user may choose another language or that the Chinese-only format is required for a region-specific or compliance-related reason, which can violate language/locale policy requirements.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The title and body instructions are written in Chinese and indicate the report should be produced in Chinese, but the file does not offer any opt-in or alternative language option. Under the policy, forcing a specific language without user choice is a natural-language policy violation unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The title and instructions are entirely in Chinese and present the guidance as mandatory reference material, but there is no indication that users may choose another language or that the skill is intentionally region-specific. This can violate a language/locale policy when a skill forces a specific language without user opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The instruction states that every storyline should use this template, and the template content is written as required Chinese text placeholders and checklist items. This imposes a specific language/locale choice without any opt-in or documented justification, which matches the policy-violation category for language constraints.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

The template title and content are written in Chinese, indicating the skill output is expected in a specific language by default. The file does not offer a user choice of language or explain a justified region-specific constraint.

Content

No source excerpt is available for this finding.