The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided "source material" and files for grammar analysis, which presents a surface for indirect prompt injection.
Ingestion points: The workflow in SKILL.md begins by inspecting "available files" and "source material."
Boundary markers: The instructions lack specific requirements for using delimiters (such as triple quotes) to isolate user-provided text from the agent's instructions.
Capability inventory: The skill has access to sandbase_run, which can perform external searches, web scraping, and model inference.
Sanitization: No specific input sanitization or instruction filtering is implemented for the source text.
[EXTERNAL_DOWNLOADS]: The skill integrates with external capabilities using the SandBase toolset (sandbase_run, sandbase_discover) to perform searches or enrichment.
Evidence: The tools are used to fetch "current external evidence" or perform "search, scraping, enrichment" as detailed in SKILL.md and references/sandbase-api-map.md.
Controls: The skill documentation mandates that the agent must present the endpoint, pricing, and total estimate to the user and obtain explicit confirmation before performing any paid calls.

grammar-check