Back to skill

Security audit

Volcengine Ai

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent Volcengine AI API helper, but it embeds a concrete API key and tells users to persist it in plaintext shell startup files.

Review before installing. Use your own scoped Volcengine API key, do not use the embedded value, and avoid storing secrets in ~/.bashrc unless you knowingly accept the exposure. Do not send confidential prompts, private image URLs, video URLs, personal data, or regulated content through this skill without an approved Volcengine data-sharing arrangement.

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:11
Finding

Hard-Coded API Credential and Insecure Plaintext Persistence

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 11–13
Vulnerability Type: Hard-coded secret and plaintext credential storage
Risk Level: High

bash
export VOLCENGINE_API_KEY="222b33d4-f22f-4f99-b68f-0eb9150ab507"
# 或在 ~/.bashrc 中持久化
echo 'export VOLCENGINE_API_KEY="222b33d4-f22f-4f99-b68f-0eb9150ab507"' >> ~/.bashrc

Technical Analysis

The Skill embeds a concrete Volcengine API credential directly in its documentation rather than using a placeholder. Anyone who can access the Skill package can extract and attempt to use this credential.

The instructions also recommend appending the credential to ~/.bashrc. This stores the secret in plaintext and automatically exports it into subsequent interactive shell environments. The credential may consequently be exposed through repository copies, workstation backups, shell configuration disclosure, diagnostic output, or access to the environments of descendant processes.

Although modifying ~/.bashrc persists a configuration value across shell sessions, the observed command does not install a backdoor or malicious startup hook. The primary issue is therefore insecure secret handling under T09, not confirmed system persistence under T06.

Attack Path

  1. An attacker obtains a copy of SKILL.md from the distributed Skill package or another location where it was copied.
  2. The attacker extracts the embedded value assigned to VOLCENGINE_API_KEY.
  3. The attacker submits requests with Authorization: Bearer <extracted-key> to the documented Volcengine Ark API.
  4. If the credential remains valid, the attacker invokes any API operations and accesses any resources authorized by its assigned permissions.
  5. Alternatively, after a user follows the persistence instruction, an attacker with read access to that user's ~/.bashrc, backups, or exported process environment can recover the same credential.

Impact Assessment

Successful exploitation could permit unauthorized use ...[truncated 485 chars]

Remediation
View remediation

Remediation Suggestions

  1. Immediately revoke the exposed credential and issue a replacement, because removing it from the current file does not invalidate copies already distributed or retained in history.
  2. Replace the concrete value with a nonfunctional placeholder:
    bash
    export VOLCENGINE_API_KEY="YOUR_VOLCENGINE_API_KEY"
    
  3. Require every user to provision an individual credential with only the minimum API permissions needed for the intended operations.
  4. Avoid recommending plaintext storage in ~/.bashrc. Prefer an operating-system credential store, secrets manager, protected CI/CD secret, or runtime credential injection.
  5. If environment variables must be used, supply them only for the required process or session and ensure they are not printed in logs or diagnostic output.
  6. Review repository history, published packages, logs, backups, and documentation mirrors for additional copies of the exposed secret.
  7. Enable provider-side usage alerts, restrictive quotas, credential expiration, and audit logging to detect and limit unauthorized use.
  8. Add automated secret scanning to the publication workflow to prevent future credentials from being committed or distributed.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

YARA rule 'backdoor_persistence': Backdoor persistence with malicious payloads (shell commands, SSH key injection, hidden root users) [malware]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · SKILL.md (reported line 17)May include surrounding context.

I能力:图片生成(Seedream-5.0-lite)、视频生成(Seedance-1.5-pro)、图片理解、视频理解。使用前需配置API密钥(VOLCENGINE_API_KEY)。支持异步任务查询。

火山引擎 AI

调用火山引擎方舟平台的AI能力。

配置

首先设置环境变量:

bash
export VOLCENGINE_API_KEY="222b33d4-f22f-4f99-b68f-0eb9150ab507"
# 或在 ~/.bashrc 中持久化
echo 'export VOLCENGINE_API_KEY="222b33d4-f22f-4f99-b68f-0eb9150ab507"' >> ~/.bashrc

模型ID

能力模型ID
图片生成doubao-seedream-5-0-lite
视频生成doubao-seedance-1-5-pro-251215
图片理解doubao-seed-1-8-251228
视频理解doubao-seed-1-8-251228

核心命令

1. 图片生成 (Seedream)

bash
# 调用火山引擎API生成图片
curl -X POST "https://ark.cn-beijing.volces.com/api/v3/visual generation/tasks" \
  -H "Authorization: Bearer $VOLCENGINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The documentation instructs users to persist an API key in ~/.bashrc and even includes a concrete key value. Persisting secrets in shell startup files increases exposure to accidental disclosure through backups, dotfile syncing, shell history mishandling, or local compromise, and it exceeds the minimum needed to demonstrate API usage. In this skill context, handling third-party API credentials is expected, but encouraging persistent plaintext storage makes the practice more dangerous than necessary.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 35)May include surrounding context.

bash
# 调用火山引擎API生成图片
curl -X POST "https://ark.cn-beijing.volces.com/api/v3/visual generation/tasks" \
  -H "Authorization: Bearer $VOLCENGINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill shows image/video URLs and prompts being sent to Volcengine endpoints without any notice that user content will leave the local system and be processed by a third party. This creates privacy and data-governance risk, especially if users provide sensitive media, internal URLs, or confidential prompts under the assumption that processing is local. Because the skill's purpose is to call an external AI service, transmission is expected, but the lack of explicit disclosure and safety guidance is still a real issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

SQP-3 covers natural-language policy violations across all file types, including forced language constraints without opt-in. The file presents all user-facing instructions and examples only in Chinese, which can impose a language choice on users without documenting that constraint or offering alternatives.

Content

No source excerpt is available for this finding.