The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The llm_strategy.py script attempts to read the local Anthropic credentials file located at ~/.claude/.credentials.json to extract an accessToken for its operations.- [DYNAMIC_EXECUTION]: The skill uses a template-based approach to generate a complete Python bot script ({name}_bot.py) at runtime, incorporating user profile data and strategy specifications.- [COMMAND_EXECUTION]: The coach.py script verifies the generated bot script by executing it using subprocess.run(['python3', filepath, '--smoke-test']), which runs locally generated code.- [COMMAND_EXECUTION]: The skill relies on the onchainos CLI tool, which is executed via subprocess.run in onchainos.py to handle all blockchain-related tasks including transactions.- [INDIRECT_PROMPT_INJECTION]: The skill's strategy generation logic in llm_strategy.py takes user-provided profile information and interpolates it into a prompt for an LLM to generate strategy code, representing an injection surface without explicit sanitization mechanisms.

starter-coach