The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection through its core functionality of retrieving and injecting persistent memories into the agent context.
Ingestion points: The recall and ask methods described in ts-sdk.md and python-sdk.md fetch memory data from the Walrus decentralized storage via a relayer.
Boundary markers: As detailed in ts-sdk.md, the withMemWal middleware automatically injects retrieved memories into the system message. The documentation does not specify the use of strict boundary delimiters or instructions to ignore embedded commands within these memories.
Capability inventory: The skill enables network communication with a relayer service and provides tools to process and reason over retrieved external data.
Sanitization: There is no explicit requirement or guidance for sanitizing or escaping the retrieved memory content before it is interpolated into the agent's prompts.

walrus-memory