The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation process requires cloning an external repository and executing a shell script (./scripts/build.sh) to build the binary. This pattern allows for the execution of arbitrary, unverified code on the host machine.\n
Evidence: The README instructs the user to run git clone https://github.com/win4r/codebase-memory-mcp-pro.git and then execute ./scripts/build.sh.\n- [EXTERNAL_DOWNLOADS]: The skill fetches its core components from a GitHub account (win4r) that is not identified as a trusted vendor or well-known service organization.\n
Evidence: Repository URL https://github.com/win4r/codebase-memory-mcp-pro.git.\n- [COMMAND_EXECUTION]: The skill provides explicit instructions to modify the local filesystem and system configurations by copying binaries to ~/.local/bin/ and modifying the Claude MCP configuration.\n
Evidence: cp build/c/codebase-memory-mcp ~/.local/bin/ and claude mcp add codebase-memory -s user -- ~/.local/bin/codebase-memory-mcp.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to index external codebases, which creates a surface for indirect injection if a processed file contains malicious graph queries or instructions targeted at the AI agent.\n
Ingestion points: The index_repository tool accepts a repo_path to ingest external code data into the agent's context.\n
Boundary markers: No specific delimiters or safety warnings for embedded content are mentioned in the tool descriptions.\n
Capability inventory: The skill includes high-privilege tools such as query_graph (executes Cypher queries), get_code_snippet (reads file contents), and explore (performs graph traversal).\n
Sanitization: There is no evidence of sanitization or validation of the indexed source code before it is processed by the graph engine.

codebase-memory-mcp-pro-knowledge-graph