T03 · Remote Payload Retrieval and Execution
- Location
README.md:55- Finding
Remote Installer Downloaded and Executed Without Integrity Verification
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill appears to do what it claims, but its setup instructions tell users to run an unverified remote installer directly in the shell.
Review the setup steps before installing. Prefer installing uv through a trusted package manager or a download-then-verify process instead of piping a remote script into `sh`; use `VOLC_API_KEY` rather than `--api-key`; avoid sending confidential prompts or private images because they are transmitted to Volcengine and web search may send prompt-derived context externally.
README.md:55Remote Installer Downloaded and Executed Without Integrity Verification
scripts/generate_image.py:3Unpinned Runtime Dependency Resolution
scripts/generate_image.py:206API Key May Be Exposed Through Command-Line Arguments
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
从 火山方舟控制台 获取 API Key Get API key from Volcengine Ark Console
设置环境变量 / Set environment variable:
The explicit curl ... | sh pattern enables arbitrary code execution from a remote source in a single command, making chaining abuse straightforward if the fetched installer is malicious or tampered with. In the context of a skill installation guide, this materially increases the chance of users executing unreviewed code on their systems.
curl -LsSf https://astral.sh/uv/install.sh | sh
The README advertises a web-search feature but does not clearly warn that prompts and possibly related request context will be transmitted to external Volcengine services and potentially search-integrated systems. Users may unknowingly include sensitive data in prompts, creating a privacy and data-handling risk even if the feature itself is legitimate.
The README documents image editing and multi-image fusion using external URLs, but it does not warn that supplied images are uploaded or otherwise transmitted to a third-party API for processing. This can expose private, copyrighted, or sensitive images to external infrastructure without sufficiently informed user consent.
The skill invokes a Python script that uses environment variables and external network access, but the manifest declares no explicit tool scope or allowed-tools boundary. That omission weakens least-privilege controls and can allow broader execution or data egress than reviewers and users would expect.
The skill advertises web search and external image URL inputs, which means user prompts, referenced images, and possibly search-derived context may be sent to third-party services. Without a clear warning and consent-oriented language, users may unknowingly expose sensitive prompts, proprietary images, or personal data to external systems.
The reference documents a web_search tool that the model may invoke based on prompt content, but it does not warn users that portions of their prompt may be transmitted to an external search capability. In an image-generation skill, prompts may contain sensitive personal, business, or unpublished information, so omitting this disclosure can lead to unintended data exposure and privacy/compliance issues.
The script sends user prompts and optional reference images directly to a third-party Volcengine API, but it does not provide an explicit user-facing notice that potentially sensitive text and images will leave the local environment. In an agent-skill context, users may assume processing is local, so private data could be transmitted externally without informed consent.
This code performs external network transmission of user-supplied prompts and optional image data to a remote API endpoint. In this skill's context, that behavior is expected for image generation, but it still creates privacy and data-governance risk if users are not clearly informed that their content will be sent off-box to a third party.
payload["tools"] = [{"type": "web_search"}]
try:
response = requests.post(GENERATE_URL, headers=headers, json=payload, timeout=300)
response.raise_for_status()
result = response.json()
The README recommends piping a remotely fetched script directly into the shell, which prevents users from inspecting the script before execution and creates a supply-chain risk if the hosting site, transport path, or installer content is compromised. This is especially risky in installation documentation because users are likely to copy-paste it verbatim.
curl -LsSf https://astral.sh/uv/install.sh | sh
The documentation suggests passing the API key via --api-key without warning that command-line arguments can be exposed through shell history, process listings, logs, or telemetry. This increases the chance of accidental credential disclosure, especially on shared systems or in recorded execution environments.
Reading credentials from environment variables is normal, but this file lacks any explicit disclosure or note that a sensitive API key will be consumed for authenticated external requests. Under the stated audit criteria, sensitive environment variable access should have some user disclosure through comments, docstrings, or user-facing help.