Back to skill

Security audit

火一五智能体扩展技能

Security checks for vulnerabilities and agentic risk

Overview

This skill is meant to copy main-agent configuration into dynamic-agent workspaces, but it can overwrite sensitive agent files and its unchecked agent ID can redirect writes outside the intended workspace.

Install only if you trust the publisher and need main-agent state copied into dynamic-agent workspaces. Treat each sync as a persistent overwrite of agent instructions, memory, tools, and skills. Do not expose agentId to untrusted input until the skill validates IDs, enforces destination containment, and adds dry-run or confirmation behavior.

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
index.ts:22
Finding

Unvalidated Agent Identifier Enables Path Traversal and Arbitrary File Overwrite

Content
View full analysis
${targetPath}`, err); } return false; } ``` ```typescript function syncDir(sourceDir: string, targetDir: string): boolean { try { if (!fs.existsSync(sourceDir)) { return false; } if (!fs.existsSync(targetDir)) { fs.mkdirSync(targetDir, { recursive: true }); } const entries = fs.readdirSync(sourceDir, { withFileTypes: true }); for (const entry of entries) { const sourcePath = path.join(sourceDir, entry.name); const targetPath = path.join(targetDir, entry.name); if (entry.isDirectory()) { syncDir(sourcePath, targetPath); } else if (entry.isFile()) { // Always prefer the main Agent version and overwrite the target file fs.copyFileSync(sourcePath, targetPath); } } return true; } catch (err) { con ...[truncated 4597 chars]
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill describes synchronization behavior without prominently warning that target files may be overwritten, even though earlier sections state the main agent's content always takes precedence. In a multi-agent workspace, this can lead users or downstream automation to destroy or replace dynamic-agent-specific configuration unexpectedly, with possible loss of isolation or operational settings.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The README first states for directory sync that conflicts are overwritten ('有冲突就覆盖') at L38, but the notes later say directories 'only copy new files' at L80. These statements are materially inconsistent about whether existing files in dynamic-agent workspaces may be replaced, which affects operator understanding of side effects and data safety.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The invocation phrase is very broad and could be matched by a generic request to sync configuration, causing the skill to trigger when the user did not intend to perform a destructive workspace synchronization. In this skill's context, that unintended trigger can overwrite configuration in a dynamic agent workspace, making prompt-based misfires operationally risky.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The file-level description and later user-facing strings are written exclusively in Chinese, with no indication that the skill is region-specific or that users can choose another language. This can violate language/locale policy when skills are expected to be language-neutral or user-selectable.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The recursive sync logic copies files from the main workspace into the dynamic agent workspace and overwrites existing files without confirmation, backup, or conflict handling. In this skill context, that is security-relevant because it can silently replace agent skills, memory, or configuration files, potentially propagating unsafe instructions or destroying integrity-critical state across agent workspaces.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The inline comment at L137 states '只复制不存在的文件' ('only copy files that do not exist'), but the called syncDir function copies every file and its own documentation at L071-L073/L092 says conflicts are overwritten. This is an active contradiction in the code documentation about whether existing dynamic-agent files are preserved.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The natural-language instructions and usage guidance are presented entirely in Chinese, which may impose a language requirement on users without opt-in. The file does not indicate that the skill is intentionally limited to a Chinese-speaking or region-specific audience.

Content

No source excerpt is available for this finding.