The Agent Skills Directory

[SAFE]: The skill is a technical guide for building and optimizing retrieval pipelines. It provides valid mathematical and logical implementations for Reciprocal Rank Fusion (RRF), Maximal Marginal Relevance (MMR), and hybrid search using BM25 and vector embeddings.\n- [PROMPT_INJECTION]: Section 6 discusses 'Contextual Compression', a technique where an LLM is used to filter retrieved document chunks. This is an architectural pattern that involves processing untrusted external data, which is a standard surface for Indirect Prompt Injection.\n
Ingestion points: The chunk_text variable in the Python compress function (SKILL.md).\n
Boundary markers: The prompt template relies on standard string interpolation without explicit escaping or advanced delimiting.\n
Capability inventory: The code performs LLM API calls via a client messenger.\n
Sanitization: No sanitization of retrieved chunks is shown, which is typical for documentation focusing on logic rather than production hardening.

retrieval-strategy