The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses the ~/.claude/ directory, which is a highly sensitive location containing agent memory (session history), agent definitions, custom commands, and MCP server configurations which may contain API keys or authentication tokens.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to run a Python module (coremind_understand) and a dashboard server using npx vite from a specific local project path.
[EXTERNAL_DOWNLOADS]: The skill has a dependency on an external, non-standard Python package named coremind_understand. Use of npx also facilitates the dynamic download and execution of the vite package.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data to build its knowledge graph.
Ingestion points: Reads files from ~/.claude/ including skills, agents, commands, and memory files.
Boundary markers: None identified; the skill directly processes the content of these files.
Capability inventory: Executes shell commands (python3, npx), reads the filesystem (~/.claude/), and launches a web server.
Sanitization: None identified; the skill appears to map and visualize the raw content of the ingested files.

understand-setup