飞书语音回复

• Outcome: Review
• Latest audit: Sep 14, 2026 · 9:36 PM UTC
• Version: 1.0.0

Overview

The skill does what it says at a high level, but it can turn chat text into audio through an external TTS tool and send Feishu messages under broad triggers without enough user-control or privacy guidance.

Review before installing. This skill is intended for Feishu voice replies, but users should avoid using it for secrets, private records, or sensitive business content unless they are comfortable sending that text to the TTS provider and into Feishu. It should be tightened to require explicit Feishu voice-message intent, document the external TTS data flow, and remove the unresolved conflict markers.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Behavioral AST: exec() Call, eval() Call, Dynamic Import
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection

Findings (7)

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 96% confidence
• Finding: The implemented code matches part of the description: it turns text into speech and converts the result into an Ogg/Opus file suitable for later use. However, the declared purpose goes further and claims the skill sends the result as a Feishu-native voice message so it appears as tappable in Feishu alongside visible text. No Feishu API calls, upload steps, message-send logic, authentication, or text-message composition are present. The code only performs local file generation and prints metadata. Therefore the description materially overstates the actual behavior.

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 90% confidence
• Finding: The skill explicitly instructs use of a shell-invoked Python script and ffmpeg, but it declares no tool scope or permissions boundary. That creates a capability gap where an agent may execute shell commands without an explicit allowlist, increasing the risk of unintended command execution or unsafe expansion of the skill's operational surface.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The invocation text includes broad triggers like voice playback or spoken summaries, which can cause the skill to activate in situations the user did not specifically intend. Over-broad activation matters here because the skill can invoke synthesis, shell-based media conversion, and outbound Feishu message behavior, expanding the chance of unnecessary processing or data being sent in audio form.

Natural-Language Policy Violations

Medium

• Confidence: 94% confidence
• Finding: The markdown sets 'Default language: Chinese' and a Chinese TTS voice, but does not state that the user can choose another language or that Chinese is only used when explicitly requested. This creates a language-policy concern because the skill appears to force a locale choice by default.

subprocess module call

Medium

• Category: Dangerous Code Execution

Content

def run(cmd):
    proc = subprocess.run(cmd, capture_output=True, text=True)
    if proc.returncode != 0:
        raise RuntimeError(proc.stderr.strip() or proc.stdout.strip() or f"command failed: {' '.join(cmd)}")
    return proc

• Confidence: 70% confidence
• Finding: subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The script sends provided text to edge-tts, which may involve transmission to an external TTS service, but it gives no disclosure, consent prompt, or guardrail about sending potentially sensitive content off-host. In a messaging skill that may synthesize arbitrary user or conversation text, this creates a real privacy and data-handling risk if secrets, personal data, or internal information are spoken via a third-party service.

Intent-Code Divergence

Low

• Confidence: 97% confidence
• Finding: The merge-conflict markers indicate the skill documentation is in an inconsistent state, leaving contradictory operational instructions in place. In agent skills, ambiguous instructions can cause unpredictable execution paths or reviewer confusion, which can mask unsafe behavior and weaken assurance around how user content is transformed before processing.

Static analysis

No suspicious patterns detected.