The Agent Skills Directory

[SAFE]: The skill operates entirely on local project files using standard Python libraries and follows the principle of least privilege. No evidence of obfuscation, remote code execution, or credential theft was found.
[PROMPT_INJECTION]: Performed analysis of the indirect prompt injection surface as the skill processes external CSV result files.
Ingestion points: The scripts/inventory_csv_results.py script reads CSV files from various project directories (e.g., results/, outputs/) to extract headers and sample data.
Boundary markers: No explicit boundary markers or instructions to ignore embedded content are used when the agent reads the CSV-derived inventory.
Capability inventory: The skill possesses Read, Write, Edit, Bash, and Glob capabilities to facilitate asset creation and provenance recording.
Sanitization: The CSV content is parsed using the standard Python csv module; the skill does not implement custom filtering for natural language instructions within data fields, but no dangerous command chains are triggered by the data ingestion.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a local, included Python script (inventory_csv_results.py). This script is used for indexing result files and does not perform network operations or access sensitive system directories.

paper-result-asset-builder