The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file contains a dynamic context command (!command) that attempts to execute a shell script named skill-toc from the plugin-utils directory at skill load time. While intended for generating documentation structures, this represents a load-time execution of external scripts.
[DYNAMIC_EXECUTION]: The script scripts/crsp_lseg_splice.py dynamically modifies sys.path to import the ds_schema module from a parent directory search. This runtime path manipulation is a medium-risk pattern for dynamic loading from computed paths.
[COMMAND_EXECUTION]: The documentation provides instructions for users to run shell commands to source credentials and execute the data processing pipeline.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process financial datasets from WRDS and LSEG APIs, which creates a potential surface for indirect injection. * Ingestion points: External database connections (WRDS) and API endpoints (LSEG RDP). * Boundary markers: None identified in the provided processing logic. * Capability inventory: Network access for API data fetching and local file system access for Parquet writes. * Sanitization: The script uses parameterized SQL queries via psycopg2, which is a best practice, but does not provide content-level sanitization for the financial data itself.

crsp-lseg-splice