codebase-memory-mcp-pro-knowledge-graph

Warn

Audited by Gen Agent Trust Hub on Jun 24, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires cloning source code from https://github.com/win4r/codebase-memory-mcp-pro.git, which is not an official or recognized trusted repository.
  • [COMMAND_EXECUTION]: The installation process involves executing ./scripts/build.sh and make on the downloaded content, running local shell scripts from an untrusted source.
  • [REMOTE_CODE_EXECUTION]: The combination of cloning an external repository and executing build scripts creates a potential remote code execution vector if the source repository contains malicious logic.
  • [DATA_EXFILTRATION]: The skill's core functions include indexing local repositories and retrieving code snippets, granting it broad read access to potentially sensitive local data.
  • [PROMPT_INJECTION]: The skill creates an indirect prompt injection surface (Category 8) because it processes untrusted codebase content that could contain instructions designed to manipulate the agent. Ingestion points: The index_repository tool processes all files within the user-specified repo_path. Boundary markers: No explicit markers are used to isolate untrusted code data from the agent's instructions. Capability inventory: The skill performs file system reads and executes complex graph queries. Sanitization: No sanitization of source code content is described before it is indexed or queried.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 24, 2026, 03:11 AM
Security Audit — agent-trust-hub — codebase-memory-mcp-pro-knowledge-graph