The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions and examples for building RAG (Retrieval-Augmented Generation) pipelines that ingest data from various untrusted sources.
Ingestion points: SimpleDirectoryReader (SKILL.md), SimpleWebPageReader (SKILL.md), GithubRepositoryReader (SKILL.md), JSONReader (SKILL.md), and DatabaseReader (SKILL.md) are documented as methods to bring external data into the agent's context.
Boundary markers: The prompt template examples, such as the qa_prompt in SKILL.md and references/query_engines.md, do not include explicit delimiters or instructions to ignore commands embedded within the retrieved context.
Capability inventory: The skill demonstrates capabilities for file system access (SimpleDirectoryReader), network requests for scraping (BeautifulSoupWebReader), and database connectivity (DatabaseReader).
Sanitization: The provided code snippets do not illustrate methods for sanitizing or validating the content retrieved from external sources before it is processed by the LLM.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions for installing necessary software components from standard, well-known package registries.
Evidence: pip install llama-index and various integration packages like llama-index-llms-openai are referenced in SKILL.md.
[COMMAND_EXECUTION]: The skill contains instructions for the user to execute shell commands for environment setup.
Evidence: Installation commands for pip and bash examples are provided in the 'Quick start' and 'Installation' sections of SKILL.md.

llamaindex