The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from a CSV file (input.csv) using the fpga_engineering_diagnostics.py script and then incorporates that summary into the agent's context. This creates a surface where instructions hidden within the data fields (e.g., in a pipeline_stage string) could attempt to influence downstream agent actions.
Ingestion points: Data is read from input.csv in scripts/fpga_engineering_diagnostics.py and then printed/saved as JSON to be used by the agent.
Boundary markers: None identified; the JSON output is processed directly by the agent without specific delimiters or instructions to ignore embedded commands.
Capability inventory: The skill uses Python scripts with pandas for data processing and writes to the local file system (--output).
Sanitization: The script performs type conversion and drops rows with invalid timestamps, but it does not sanitize string content against potential prompt injection attacks.

fpga-engineering