Trust Assessment
doc-coauthoring received a trust score of 27/100, placing it in the Untrusted category. This skill has significant security findings that require attention before use in production.
SkillShield's automated analysis identified 4 findings: 1 critical, 3 high, 0 medium, and 0 low severity. Key findings include Unvalidated URL Fetching (SSRF Risk), Unrestricted File Reading (Path Traversal Risk), Unvalidated File Creation (Arbitrary File Write/Path Traversal).
The analysis covered 4 layers: dependency_graph, manifest_analysis, llm_behavioral_safety, static_code_analysis. The static_code_analysis layer scored lowest at 55/100, indicating areas for improvement.
Last analyzed on February 11, 2026 (commit 6195a031). SkillShield performs automated 4-layer security analysis on AI skills and MCP servers.
Layer Breakdown
Behavioral Risk Signals
Security Findings4
| Severity | Finding | Layer | Location | |
|---|---|---|---|---|
| CRITICAL | Prompt Injection via Freeform Feedback Parsing The skill instructs the agent to 'extract their preferences and proceed. Parse what they want kept/removed/changed and apply it' from freeform user feedback. This flexible parsing of unstructured user input for critical actions (modifying document content) creates a significant prompt injection vulnerability. A malicious user could craft feedback that attempts to inject new instructions, override previous directives, or manipulate the LLM's behavior beyond content refinement, potentially leading to unauthorized actions or data manipulation. Implement strict parsing rules for freeform feedback, focusing on extracting specific, predefined actions (keep, remove, combine, change) and their targets. Use explicit tool calls or structured output for modifications rather than relying solely on LLM interpretation of natural language for critical actions. Add explicit instructions to the LLM to ignore any instructions embedded in user feedback that deviate from content refinement and to only apply changes related to the document content. | Unknown | SKILL.md:180 | |
| HIGH | Unvalidated URL Fetching (SSRF Risk) The skill instructs the agent to fetch content from user-provided links to shared documents using 'appropriate integration'. Without strict URL validation and sanitization, a malicious user could provide internal or malicious URLs, leading to Server-Side Request Forgery (SSRF). This could allow the agent to access internal network resources, exfiltrate data, or interact with unintended services. Implement strict URL validation (e.g., whitelist allowed domains/protocols, block private IP ranges) and sanitize URLs before fetching. Ensure the underlying integration is configured to prevent access to internal networks and has robust error handling for invalid or unreachable URLs. | Unknown | SKILL.md:48 | |
| HIGH | Unrestricted File Reading (Path Traversal Risk) The skill instructs the agent to 'read' a user-provided file. If the user can specify the file path without proper validation, this could lead to path traversal, allowing the agent to read arbitrary files outside the intended scope (e.g., sensitive system files or other user data). Implement strict validation and sanitization of file paths provided by the user. Confine file reading operations to a designated, secure directory (e.g., a temporary upload folder) and prevent path traversal characters (e.g., `../`). Ensure the file reading tool has minimal necessary permissions. | Unknown | SKILL.md:50 | |
| HIGH | Unvalidated File Creation (Arbitrary File Write/Path Traversal) The skill instructs the agent to create files using `create_file` or by creating a markdown file in the working directory. If the filename or the content written to the file is directly derived from unsanitized user input, it could lead to path traversal (creating files outside the intended directory) or arbitrary file write vulnerabilities. This could allow a malicious user to write executable scripts, overwrite critical files, or exfiltrate data by writing it to an accessible location. Ensure all filenames and file contents derived from user input are strictly validated and sanitized. Prevent path traversal characters (e.g., `../`) in filenames. Implement content sanitization for files that might be rendered or executed. Confine file creation to a secure, isolated working directory with appropriate permissions. | Unknown | SKILL.md:137 |
Scan History
Embed Code
[](https://skillshield.io/report/8caaaf4b79bccc91)
Powered by SkillShield