The Agent Skills Directory

[SAFE]: The skill serves as a reference for developers and does not contain any executable malicious payloads or unsafe commands.
[INDIRECT_PROMPT_INJECTION]: The skill documents patterns for RAG and tool usage that process external data, which creates a potential surface for indirect prompt injection in systems built using these templates.
Ingestion points: External document content is ingested in the retrieve_and_generate function in ai-ml-integration.md and via the url argument in the summarize prompt handler in mcp-patterns.md.
Boundary markers: Code examples use standard text delimiters such as "Context:" and "Question:" to separate external data from instructions.
Capability inventory: The provided patterns include tools for file system access (read_file), database interaction (search_database), and network operations (send_email).
Sanitization: The skill includes a 'Security Considerations' checklist in SKILL.md that explicitly advises users to validate input parameters and sanitize file paths.

mcp-development