The Agent Skills Directory

[SAFE]: The skill's functionality is consistent with its role as an academic research designer, focusing on methodological rigor and ethical standards.
[DATA_EXFILTRATION]: The skill accesses project-related files (topic proposals, literature reviews) to provide context for research design. No sensitive system files or credentials are accessed, and no unauthorized network transmissions were found.
[PROMPT_INJECTION]: The instructions do not contain safety bypasses or attempts to manipulate agent behavior. It uses instructional language to define its role and quality standards.
[PROMPT_INJECTION]: (Indirect Surface) The skill processes user-supplied research documents and has the capability to write files and search the web, which is an inherent surface for indirect prompt injection. However, it lacks high-privilege tools or dangerous commands that would allow for significant exploitation.
Ingestion points: Reads existing project files (topic proposals, literature reviews, advisor feedback) in the Workflow section.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present.
Capability inventory: Web search, file read, and file write ({project}-research-design.md).
Sanitization: No evidence of output sanitization or validation of ingested content.

alterlab-rma-research-designer