content-creator
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on local Python and Shell scripts (
extract-template.py,create-content.py,setup-project.sh) to implement its core workflow. These scripts are invoked by the agent to analyze writing styles, generate content, and manage project structures.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided articles and web search results to extract writing styles and gather information. The lack of explicit boundary markers or sanitization when processing this external content creates a surface for indirect prompt injection.\n - Ingestion points: Sample articles (SKILL.md Step 1) and external research materials (SKILL.md Step 4).\n
- Boundary markers: None identified in the prompt instructions.\n
- Capability inventory: Local script execution, file system access, and note-taking tool integration.\n
- Sanitization: None identified.\n- [DATA_EXFILTRATION]: The
archive-project.pyscript allows for the copying of arbitrary files and directories into a project's archive folder. If an agent is manipulated into archiving sensitive system files (e.g., credentials or configuration files), it could lead to unintended data exposure.
Audit Metadata