The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a Retrieval-Augmented Generation (RAG) pattern, which introduces a surface for indirect prompt injection where untrusted content from the database is processed by the agent.
Ingestion points: External research data is retrieved from the harness_embeddings table in Neon via SQL queries.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the retrieved text chunks.
Capability inventory: The skill allows the agent to execute SQL queries and read local markdown files identified in search results.
Sanitization: No explicit sanitization or validation of the retrieved content is mentioned before it is processed by the LLM.
[SAFE]: The skill contains specific security guidance that aligns with best practices, such as preventing the logging of sensitive environment variables like DATABASE_URL or API keys. It also enforces data isolation by requiring a WHERE clause to scope queries to a specific corpus_id.

research-rag