marketing-science-academic-writing
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides example shell commands (
cp -r) for the user to manually copy skill files into specific local agent configuration directories (e.g.,~/.claude/skills/). These are standard installation instructions for local agent skills. - [DYNAMIC_EXECUTION]: The skill generates Python code snippets using standard scientific libraries (
pyblp,numpy,pandas,statsmodels) and LaTeX templates for manuscript assembly. This behavior is the primary intended function of the writing assistant and is performed on the user's local environment with standard tools. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a template for processing research ideas and data descriptions provided by the user. While it ingests external text, it lacks any high-privilege capabilities (like network exfiltration or system modification) that would make it a viable target for injection attacks.
Audit Metadata