The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes product information from external web sources, which is a standard surface for indirect prompt injection attacks where malicious instructions could be embedded in web content.
Ingestion points: External product data and specifications retrieved via WebSearch and WebFetch tools during the search phase (Step 4).
Boundary markers: The instructions do not specify explicit delimiters for the external content, though the results are presented to the user in a structured Markdown format.
Capability inventory: The skill can execute a local Python script using the Bash tool and perform project file operations via the Write tool (Step 6).
Sanitization: The risk is mitigated by a mandatory human-in-the-loop confirmation gate (AskUserQuestion) which requires the user to approve the pairings and the target file path before the saving script is executed.
[COMMAND_EXECUTION]: The skill executes a localized Python script to perform structured updates to the project's product library.
Evidence: python3 "<plugin-root>/skills/master-schedule/scripts/csv-library.py" append product --project <project-root> --row-json <batch.json>
Context: This command invokes a script residing within the plugin's own directory structure (<plugin-root>). As the script is a resource provided by the skill's author (alpacalabsllc), this constitutes expected functionality for managing project data.

product-pair