The Agent Skills Directory

[SAFE]: Analysis of the instructions and provided rulesets confirms that the skill does not use obfuscation, hardcoded credentials, or unauthorized external connections. All tools are vendor-specific or standard for graph database interaction.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because the agent is instructed to read data from an external source (the database) and interpret it.
Ingestion points: Data entering the context from execute_cypher, search_nodes, vector_search, and text_search in SKILL.md.
Boundary markers: The instructions do not establish explicit delimiters or include 'ignore instructions' warnings for content retrieved from tool outputs.
Capability inventory: The agent has the ability to execute arbitrary Cypher queries (read/write/schema mutation) through execute_cypher and can import bulk data via import_rdf.
Sanitization: There are no instructions for sanitizing or validating retrieved data before it is used by the agent to make decisions or generate responses.

ogdb-cypher