The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill contains a self-referential assertion regarding its own verification status. This takes the form of a comment indicating that a verification script produces a 'PASS' result, which is a pattern used to influence the perception of the skill's safety.
Ingestion points: The skill processes Channel State Information (CSI) sensor data from local directories such as data/csi/, data/mmfi/, and data/paired/ (SKILL.md).
Boundary markers: None mentioned or enforced in the provided instructions.
Capability inventory: The skill has extensive system access, including executing shell scripts, running compiled Rust binaries (cargo run), and interacting with cloud CLI tools (gcloud) (SKILL.md).
Sanitization: No sanitization or validation of the input sensor data is described.
Evidence: python archive/v1/data/proof/verify.py # VERDICT: PASS (SKILL.md)
[COMMAND_EXECUTION]: The skill provides instructions to execute numerous local scripts and binaries across various languages, including Python, Node.js, and Rust. It also automates authentication and configuration for external cloud platforms.
Evidence: gcloud auth login, bash scripts/gcloud-train.sh, python scripts/publish-huggingface.py, and multiple cargo run commands (SKILL.md).

ruview-model-training