The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied BIM/CAD data in various formats (CSV, Excel, JSON, IFC) which creates a surface for indirect injection.
Ingestion points: The skill uses pd.read_csv, pd.read_excel, and ifcopenshell.open in SKILL.md to load external datasets.
Boundary markers: While the instructions in instructions.md direct the agent to validate inputs, there are no specific prompt delimiters or instructions to ignore commands embedded within the model data.
Capability inventory: The skill utilizes filesystem write capabilities (to_excel, to_csv, to_json) and data processing libraries.
Sanitization: There is no explicit logic for sanitizing or escaping string data extracted from BIM models before it is presented or processed further.
[EXTERNAL_DOWNLOADS]: The skill references external resources for documentation and tooling.
It links to the vendor's website datadrivenconstruction.io for methodology and book references.
It references ifcopenshell.org for BIM data extraction utilities.
[COMMAND_EXECUTION]: The skill's configuration in SKILL.md specifies requirements for system binaries including python3 and the ifcopenshell environment.

qto-report