The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Fetches and executes the uv package manager installer from the official domain of a well-known technology provider (astral.sh).
[COMMAND_EXECUTION]: Executes an embedded Python script to initialize and update the local preference file at ~/.ouroboros/prefs.json.
[COMMAND_EXECUTION]: Modifies the agent's global configuration file (~/.claude/mcp.json) to register the Ouroboros MCP server, enabling persistent tool access across sessions.
[DATA_EXFILTRATION]: Invokes the GitHub CLI (gh api) to perform an authenticated outbound action (starring a repository) on behalf of the user.
[PROMPT_INJECTION]: Indirect prompt injection surface exists as the skill processes and modifies user-controlled project and configuration files.
Ingestion points: Reads content from CLAUDE.md and system-level configuration files (~/.claude/mcp.json).
Boundary markers: No specific markers or delimiters are used to isolate processed file content from internal agent instructions.
Capability inventory: The skill possesses capabilities including shell command execution (curl, python3, gh, uvx, pipx), file system modification, and authenticated network operations.
Sanitization: The skill uses standard JSON parsing for configuration data but lacks content validation or sanitization for markdown file modifications.

setup