The Agent Skills Directory

[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection in modules/research-checklist.md by instructing the agent to perform web research and incorporate findings from external sources.
Ingestion points: External data is ingested through web searches and external tools mentioned in modules/research-checklist.md (e.g., Wikipedia and Arxiv distillers).
Boundary markers: There are no defined delimiters or instructions to treat external data as untrusted text within the workflow.
Capability inventory: The agent is directed to use research findings to "Update metadata" and "Update modules" (including modules/core-guidance.md), which involves writing or modifying its own configuration and instructions.
Sanitization: No sanitization or validation mechanisms are specified for the content retrieved from external sources before it is processed.

oceanography-specialist