The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes a workflow for processing external data, which introduces a potential attack surface if the input data contains adversarial instructions.
Ingestion points: The skill ingests untrusted data from the data/ directory and project target paths during feature engineering and experimentation.
Boundary markers: There are no explicit instructions or delimiters provided to help the agent distinguish between its system instructions and potential instructions embedded within the data files.
Capability inventory: The agent is authorized to execute complex operations including running local Python scripts, building Docker images, and managing production environments via kubectl and helm.
Sanitization: The skill does not specify any input validation, sanitization, or safety checks for the data being processed.
[COMMAND_EXECUTION]: The skill provides several command-line instructions for the agent to perform data science and DevOps tasks.
Evidence: The instructions include commands for local script execution (python scripts/experiment_designer.py), containerization (docker build), and orchestration (kubectl apply, helm upgrade). While standard for a senior data scientist role, these commands grant the agent broad control over the local system and connected infrastructure.

senior-data-scientist