The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation recommends installing the @orama/orama package via npm. This is a well-known, legitimate search engine library used for high-performance full-text and vector search.
[COMMAND_EXECUTION]: The core API examples demonstrate the use of the Node.js fs module for reading and writing index snapshots to the local file system (e.g., index.orama.json). These operations are necessary for the skill's stated purpose of local persistence.
[PROMPT_INJECTION]: The implementation of a RAG pipeline inherently creates an indirect prompt injection surface where untrusted data from local files could influence the agent's behavior.
Ingestion points: Untrusted data enters the system through the indexFile function in SKILL.md, which processes file content for indexing.
Boundary markers: The provided code snippets do not implement specific boundary markers or instructions for the agent to ignore embedded commands within retrieved context.
Capability inventory: The skill enables the agent to search and retrieve text segments from local files to serve as context for its responses.
Sanitization: There is no visible sanitization or filtering of content during the indexing or retrieval phases.
Assessment: This vulnerability surface is typical for RAG systems and is considered acceptable given the skill's primary purpose of providing local semantic search capabilities.

rag