The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions direct the agent to access files located at ../../resources/source-basis.md and ../../resources/official-source-map.md. This represents a directory traversal pattern where the agent is prompted to read files outside of the skill's specific directory. While these files appear to be internal resources for the platform, accessing files beyond the skill's scope can lead to unauthorized data exposure.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to ingest and act upon untrusted external data, including user-provided manuscripts and live web search results for journal instructions.
Ingestion points: User-submitted manuscript content and external web pages retrieved via search tools (e.g., "Search the live site for 'Science Advances author instructions'").
Boundary markers: The skill does not provide delimiters or instructions to the agent to ignore potentially malicious commands embedded in these external sources.
Capability inventory: The agent is prompted to read local files via relative paths and perform web searches.
Sanitization: No evidence of input validation, escaping, or sanitization of external content is present.

science-advances