The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download and install a Python package directly from a URL targeting a non-standard domain.
Evidence: python -m pip install ... "tuplatform-connect @ https://connect.aiscientist.tools/downloads/tuplatform_connect-0.3.0-py3-none-any.whl#sha256=3fad5eee5ecf7887a693d93ccd1aa112dc0955617a885d1fc3daded0030f9ae0" in SKILL.md.
Note: Although a SHA-256 hash is provided for integrity, the source domain is not a well-known service or a recognized trusted organization.
[COMMAND_EXECUTION]: The skill requires the execution of various shell commands and scripts to initialize the environment and validate the setup.
Evidence: python scripts/remote_validation/setup_skill_preflight.py --implementation uspto-downloader and various tu remote CLI commands in SKILL.md.
[DYNAMIC_EXECUTION]: The instructions include heredoc patterns to dynamically generate and execute Python scripts for local verification.
Evidence: Multiple blocks using `python
<<'PY'to list tools and call the MCP server inSKILL.md`.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external patent data, which creates a surface for indirect prompt injection if the retrieved content contains malicious instructions.
Ingestion points: USPTO HTTPS hosts (specified in SKILL.md).
Boundary markers: None explicitly defined in the prompt templates (absent).
Capability inventory: Network access to USPTO, local file system writes to caches/ and runs/, and local server hosting via fastmcp.
Sanitization: The skill advises the user to "Check... absence of paths, secrets, and traces" but does not describe automated sanitization or escaping mechanisms.

setup-uspto-downloader-remote-tool