arkweb-code-analysis

Warn

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill configures a remote MCP (Model Context Protocol) endpoint at https://mcp.deepwiki.com/mcp. This connects the agent to an external service for repository indexing and AI-based querying, which acts as an external data and tool provider.
  • [COMMAND_EXECUTION]: The skill instructs the agent to use git clone --depth=1 on repositories hosted at gitcode.com. While intended for source code analysis, cloning from external sources involves executing shell commands to fetch potentially untrusted content.
  • [PROMPT_INJECTION]: The skill processes data from external sources including the DeepWiki API and remote Git repositories. This creates a surface for indirect prompt injection where malicious instructions embedded in documentation or code could influence the agent's behavior.
  • Ingestion points: DeepWiki API tool outputs (read_wiki_contents, ask_question), Git repository source code, and browser snapshots from DeepWiki web pages.
  • Boundary markers: Absent. There are no instructions provided to the agent to treat external content as data only or to ignore embedded instructions.
  • Capability inventory: The skill has capabilities for shell command execution (git clone), file system writes to {DOCS_REPO}, and remote network calls via MCP and browser tools.
  • Sanitization: Absent. The skill does not define any filtering, validation, or escaping mechanisms for the data retrieved from external repositories or analysis services.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 8, 2026, 09:00 AM
Security Audit — agent-trust-hub — arkweb-code-analysis