The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates long-term memory persistence for AI agents, creating a surface where malicious instructions could be stored and subsequently executed during future interactions.
Ingestion points: Data enters the memory system through method calls such as ctx.memory.add and ctx.memory.collection().set documented in SKILL.md.
Boundary markers: The skill does not provide or mandate the use of boundary markers or instructions to the LLM to ignore embedded commands within retrieved memories.
Capability inventory: The skill enables project-wide, contact-specific, and conversation-scoped data storage and retrieval. While the skill itself does not perform dangerous operations, the retrieved data influences the agent's reasoning.
Sanitization: No explicit sanitization, filtering, or validation of the stored memory text is implemented or described in the provided examples.
[SAFE]: All external interactions and dependencies are managed through official vendor infrastructure belonging to the author (zavudev).
The use of the npx zavudev CLI tool and the @zavudev/functions package represents standard usage of vendor-provided developer tools.
Network requests are directed to api.zavu.dev, which is the authorized API domain for the service.
Credential management uses the $ZAVU_API_KEY environment variable, which is a secure and standard practice for API authentication.

memory