The Agent Skills Directory

[SAFE]: The skill provides technical documentation for internal Rust logic for text processing and embedding generation. No malicious behavior or unsafe commands were detected.
[EXTERNAL_DOWNLOADS]: The documentation notes that build-time dependencies (ONNX Runtime) are fetched via standard build processes (ort-bundled). This is a routine operation for machine learning libraries and uses well-known sources.
[INDIRECT_PROMPT_INJECTION]: The skill defines a data ingestion surface by processing external text for chunking. * Ingestion points: chunking::chunk_text and chunking::rag::chunk_for_rag in crates/xberg/src/chunking/. * Boundary markers: Structural splitting based on character counts, markdown headings, or semantic topic shifts. * Capability inventory: Embedding generation via ONNX or static models; no file-write or network operations are defined in these instructions. * Sanitization: None specified; input is split structurally based on configuration.

chunking-embeddings