The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches the 'uv' package manager via a remote script from Astral (astral.sh). Astral is a well-known service in the Python development ecosystem, and this setup is a standard deployment pattern. It also clones the 'autoresearch' repository from a well-known public researcher profile on GitHub.
[COMMAND_EXECUTION]: The skill utilizes several Bash scripts to automate the research workflow, including hardware verification (check-hardware.sh), environment setup (setup.sh), and the main autonomous loop (run-loop.sh). These scripts execute localized operations such as checking CUDA availability, syncing dependencies, and managing git commits for experiment ratcheting.
[REMOTE_CODE_EXECUTION]: While the skill contains patterns to download and execute the Astral installation script, this is handled as a safe operation per the trusted service guidelines. The primary function of the skill involves the agent modifying and executing a local 'train.py' script; this dynamic execution is the core purpose of the ML research workflow and is confined to the local repository context.
[PROMPT_INJECTION]: The skill's instructions focus on operational modes and routing. It contains no attempts to bypass safety filters or override system-level constraints. It explicitly includes 'route-out' logic to direct non-ML tasks (like general skill optimization) to appropriate alternative tools, preventing scope creep and potential instruction confusion.

autoresearch