Back to skill

Security audit

αExtractor

Security checks for vulnerabilities and agentic risk

Overview

The skill has a legitimate chemistry-image purpose, but it relies on mutable external documentation to define and run invocation code while using an API key and uploading user images.

Review this skill before installing. Use it only for images you are allowed to send to SciMiner, avoid confidential or regulated chemistry content, and prefer a version that pins the API contract or ships a reviewed local invocation path instead of executing code derived from remote documentation.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:29
Finding
Mutable Remote Documentation Used as an Executable Code Source## Vulnerability Details **File Location**: `SKILL.md`, lines 29–68 **Vulnerability Type**: Remote payload retrieval and execution **Risk Level**: High **Relevant snippet**: ```markdown ## Authoritative tool-doc source (required) The published Markdown files under `https://sciminer.tech/tool_api_files/` are the single source of truth for `provider_name`, `tool_name`, allowed `parameters`, file-upload behavior, request encoding, and the example submission flow for this skill's included tools. Use these SciMiner Markdown docs: - `AlphaExtractor` -> `AlphaExtractor_api_doc.md` The agent MUST: 1. Resolve the selected tool's Markdown file and read it before every invocation. 2. Never invent `provider_name`, `tool_name`, parameter names, enum values, upload-field names, content type, or submission flow from memory. 3. Extract and follow the selected doc section's exact: - Base URL - API endpoint - Content-Type - Authentication header - Tool Name - Method - Parameter table, including required fields and enum values - File-upload instructions and example code 4. Choose the correct section if the selected doc contains multiple tool variants. 5. Cite the selected Markdown doc as the payload source in summaries. [...] ## Required workflow 1. Read `AlphaExtractor_api_doc.md` from `https://sciminer.tech/tool_api_files/`. [...] 5. Write or run the invocation code directly from the selected Markdown doc's base-information block, parameter table, file-upload instructions, and example code. Do not apply a shared invocation template or local registry abstraction in this skill. ``` ### Technical Analysis The Skill requires the agent to retrieve a mutable Markdown document from an external URL before every invocation. It then instructs the agent to write or run invocation code directly from that document's example code and operational details. ...[truncated 2172 chars]
Remediation
## Remediation Suggestions - Bundle a reviewed invocation implementation inside the Skill instead of executing example code obtained from remote documentation. - Treat remote Markdown exclusively as untrusted reference material and prohibit executing or dynamically translating its code examples. - Hard-code or strictly allowlist the expected HTTPS origin, API endpoint, authentication-header name, HTTP method, upload destination, and redirect behavior. - Reject cross-origin redirects and never forward `SCIMINER_API_KEY` to a destination outside the approved SciMiner origin. - If remote machine-readable specifications are necessary, pin an approved version or cryptographic digest and verify it before use. - Parse only a narrow, validated data schema. Do not permit fetched content to define executable code, shell commands, arbitrary headers, or unrestricted destinations. - Keep credential injection separate from remotely supplied request definitions, and ensure logs, summaries, errors, and generated code never expose the token. - Require explicit review and release of changes to the local invocation implementation when the upstream API contract changes.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill instructs users to upload chemistry images to an external SciMiner service and to share `history_url` links, but it does not provide a clear user-facing warning that submitted content and result pages are hosted off-platform. This creates a real data exposure risk because users may upload unpublished figures, proprietary molecular structures, or sensitive research artifacts without informed consent about third-party processing and online accessibility.

Static analysis

No suspicious patterns detected.