The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/wrapper_resolver.py script contains a Bash resolver logic (RESOLVER_SNIPPET) that is embedded into the generated superra wrapper and task-hook shim. This snippet includes a fallback mechanism that attempts to shallow-clone a repository from GitHub (https://github.com/FuZhiyu/superRA.git) if the tool's scripts cannot be found locally on the system. This involves using git clone and git fetch at runtime.
[REMOTE_CODE_EXECUTION]: The CLI wrapper generated by scripts/wrapper_resolver.py includes logic to execute the scripts downloaded via the GitHub fallback mechanism. It uses exec uv run --script or python3 on paths within the cloned repository, which constitutes execution of remote code.
[COMMAND_EXECUTION]: The skill makes extensive use of the subprocess module to interact with the system environment.
scripts/plan_dashboard.py uses subprocess.Popen to spawn the dashboard server as a background process and subprocess.run to execute git commands for repository information and process management.
scripts/task_hook.py, scripts/_comments.py, and scripts/_worktree_discovery.py use subprocess.run to execute git commands for configuration, logging, and worktree management.
[DATA_EXFILTRATION]: The scripts/plan_dashboard.py script implements a web server using FastAPI that provides a /files/{path:path} endpoint. This endpoint allows reading files from the project root. While it includes a check to prevent path traversal outside the project root using is_relative_to, it provides a mechanism for accessing and potentially exfiltrating project files if the server is exposed to the network (e.g., via the --host 0.0.0.0 option).
[SAFE]: The skill vendors several well-known third-party JavaScript libraries (DOMPurify, KaTeX, highlight.js, markdown-it) for its dashboard. These are stored locally in the scripts/vendor/ directory and are used for rendering content in the standalone export mode.

task-tree