The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a retrieval-augmented generation (RAG) pattern that is vulnerable to indirect prompt injection due to the lack of sanitization of indexed content.
Ingestion points: Document content enters the agent's context through the searchDocuments and indexDocuments functions in rag-search.js, which read files from the directory specified by the user-provided --path argument.
Boundary markers: The skill output contains raw document chunks without delimiters, boundary markers, or specific instructions to the agent to treat the retrieved text as untrusted data.
Capability inventory: The script performs file system operations including directory traversal (fs.readdirSync), reading files (fs.readFileSync), and writing the search index to the local user directory (fs.writeFileSync). It also defines a class for network requests to OpenAI's embeddings API, although this is currently unused in the main execution flow.
Sanitization: No sanitization, escaping, or validation is performed on the content of the indexed documents before the text is returned as a search result.

rag-search