The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from multiple sources including user-pasted notes, CSV files, and conversation history.
Ingestion points: Raw product data provided as described in README.md and SKILL.md ("Normalize supplied notes, CSV, pasted lists, observations or selected workbook data").
Boundary markers: The instructions lack explicit directives for the agent to treat input data as strictly untrusted or to use specific delimiters to prevent potential instruction injection from the content of the data being imported.
Capability inventory: The skill has access to a broad set of tools including Write, Edit, and Bash, which could potentially be targeted if the agent obeys instructions embedded in the imported product data.
Sanitization: The implementation leverages structured operations such as product_observations.validate and ffe_intake.prepare to enforce data shapes and uses logical hashing for manifest integrity, providing a layer of structural validation.

product-data-import