The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the use of the eve db sql command with the --write flag to allow agents to interact directly with managed Postgres databases for structured memory. This is a primary platform feature for state management.
[PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by teaching agents to ingest data from shared sources. Ingestion points: .eve/coordination-inbox.md, eve docs read, eve search, eve thread messages, and eve job attachment. Boundary markers: Not present in the provided examples. Capability inventory: eve db sql (SQL execution), eve job create (task dispatch), and eve docs write (persistent storage). Sanitization: No specific input sanitization or filtering logic is demonstrated for external data.
[DATA_EXFILTRATION]: The skill details methods for writing agent discoveries and knowledge to organization-wide storage via eve docs write and eve memory set. It also describes using eve fs sync for bidirectional file synchronization between local environments and organization storage, which are standard platform functions for knowledge persistence.

eve-agent-memory