The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill provides code examples for LangChain integration in SKILL.md that use the parameter allow_dangerous_deserialization=True. This flag explicitly enables the use of the Python pickle module when loading vector indices from the local filesystem. Loading a maliciously crafted index file could lead to arbitrary code execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed for RAG (Retrieval-Augmented Generation) workflows, which are susceptible to indirect prompt injection if the retrieved content is not properly handled.
Ingestion points: External data is ingested into the FAISS index via index.add(vectors) and FAISS.from_documents(docs, ...) as shown in SKILL.md.
Boundary markers: The provided implementation examples do not include boundary markers (e.g., XML tags or delimiters) or instructions to ignore instructions embedded within the retrieved data.
Capability inventory: The skill enables reading and writing files to the local disk (faiss.write_index, faiss.read_index, vectorstore.save_local, vectorstore.load_local) and performing high-performance vector retrieval.
Sanitization: There is no evidence of sanitization or validation of retrieved content before it is passed to the LLM context.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install common, well-known libraries such as faiss-cpu, faiss-gpu, and numpy from standard package registries.

faiss