The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the bun runtime to execute commands for saving and retrieving memories. A shell script scripts/memories.sh acts as a wrapper, safely passing arguments to the underlying Bun tasks.
[DATA_EXFILTRATION]: Configuration is managed through environment variables such as OPENAI_API_BASE_URL. The skill references internal Kubernetes service URLs for its embedding model API, and no evidence of data exfiltration or credential exposure was found.
[PROMPT_INJECTION]: The skill handles arbitrary text content which is stored as embeddings, creating a potential surface for indirect prompt injection.
Ingestion points: Input content and summaries stored via the save-memory task.
Boundary markers: No explicit delimiters or guardrails are defined in the provided script wrappers.
Capability inventory: The skill can perform database operations and network requests to an embedding service.
Sanitization: No sanitization of stored content is observed in the provided shell and JSON files.

memories