The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user queries to apply financial models, which presents a surface for indirect prompt injection.
Ingestion points: User-provided investment questions, such as the example query "AI stocks are ripping and I'm worried about a bubble."
Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within user data.
Capability inventory: The skill's capabilities are restricted to reading local markdown files and generating text; it has no access to shell commands, network writing, or sensitive environment variables.
Sanitization: No input filtering or escaping mechanisms are specified for the ingested user content.
[EXTERNAL_DOWNLOADS]: The skill references various external URLs to provide source documentation for its investment frameworks.
Evidence: Links to berkshirehathaway.com, fortune.com, nytimes.com, cnbc.com, and columbia.edu are included in the article-index.md and reference files.
Context: These references target official repositories, well-known news organizations, and academic institutions. The links are used to ground analysis in authoritative sources and do not involve runtime code execution.

investor-warren-buffett