The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes research reports and their cited sources, creating an attack surface for indirect prompt injection where malicious instructions in a report could influence agent behavior.
Ingestion points: Content from .plans/research/*/REPORT.md and content from cited files or external URLs.
Boundary markers: The instructions do not define explicit delimiters or warnings to ignore instructions within the ingested content.
Capability inventory: The skill has the ability to read arbitrary files via Read and access the internet via WebFetch based on citations.
Sanitization: No explicit sanitization or validation of the report content or citation content is performed prior to processing.
[EXTERNAL_DOWNLOADS]: The skill utilizes the WebFetch tool to retrieve content from arbitrary URLs found in topic research reports. This involves making network requests to non-whitelisted external domains based on data provided in report files.

deep-study