The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads an external Python wheel for the ToolUniverse Connect relay. Evidence: pip installation of 'tuplatform-connect' from 'https://connect.aiscientist.tools/downloads/tuplatform_connect-0.3.0-py3-none-any.whl'. Security Control: The skill includes a SHA-256 integrity hash (3fad5eee5ecf7887a693d93ccd1aa112dc0955617a885d1fc3daded0030f9ae0) in the install command to verify the package hasn't been tampered with.
[COMMAND_EXECUTION]: The skill executes local preflight and validation scripts to audit the environment. Evidence: 'python scripts/remote_validation/setup_skill_preflight.py --implementation paga' in SKILL.md. Context: These commands are used for environment validation and implementation auditing, which are consistent with the primary purpose of a setup skill.
[INDIRECT_PROMPT_INJECTION]: The skill processes external datasets (.h5ad files) which represents a surface for indirect instructions. 1. Ingestion points: 'adata_path' parameter in SKILL.md. 2. Boundary markers: The instructions explicitly warn to 'Check scientific meaning, output bounds, invalid-input behavior, and absence of paths, secrets, and traces'. 3. Capability inventory: The 'run_paga_trajectory' tool is used for data analysis. 4. Sanitization: The documentation recommends manual verification of tool outputs to ensure no traces or secrets are leaked.

setup-paga-remote-tool