The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents patterns for Retrieval-Augmented Generation (RAG) where retrieved database content is directly interpolated into prompts for Large Language Models.
Ingestion points: Search result content from entities such as documents, contacts, and notes as processed in references/search-rag.md.
Boundary markers: Simple instructional delimiters (e.g., Context:) are present in references/search-rag.md, but these provide minimal protection against malicious instructions embedded in the retrieved data.
Capability inventory: The skill enables broad database read operations and integrates with LLM interaction tools via agentic-kit as seen in references/search-rag.md.
Sanitization: No explicit sanitization, filtering, or validation of the retrieved content is shown before prompt interpolation.
[EXTERNAL_DOWNLOADS]: The skill references the vendor's official code repository for related skill configurations.
Evidence: A link to https://github.com/constructive-io/constructive/tree/main/.agents/skills/pgpm is provided in references/search-postgis.md. This targets the author's own official resource.

constructive-search