article-writing
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process security findings and evidence data (e.g., from
memory/orevidence/*.json) to generate reports. This constitutes an attack surface where malicious instructions embedded in findings could influence the agent's behavior during report generation. - Ingestion points: Processes files matching
memory/*target*.mdand JSON files inevidence/directories across multiple scripts. - Boundary markers: Absent in the automation scripts; the skill relies on the agent following manual instructions for report structure.
- Capability inventory: Extensive use of
Bashfor command execution,Writefor file creation, andWebFetchfor network verification and NVD lookups. - Sanitization: The skill provides manual checklists and instructions for sanitizing reports, but the included automation scripts (e.g., in
payloads.md) do not implement strict input validation before processing findings. - [EXTERNAL_DOWNLOADS]: The skill references and provides installation commands for common Python libraries and system utilities required for report generation.
- Python packages: Instructions mention installing
jinja2andcvssfrom PyPI. - System utilities: Scripts use
pandoc,xelatex,chromium(headless),exiftool,jq, andimagemagick(convert). - Remote lookups: The
Batch CVSS Calculatorscript fetches vulnerability data from the National Vulnerability Database (NVD) viahttps://services.nvd.nist.gov, which is a well-known and trusted service. - [COMMAND_EXECUTION]: The skill relies on extensive use of shell and Python scripts for data processing, markdown conversion, and evidence formatting. These include scripts for calculating CVSS scores, generating executive summaries, and converting findings between JSON and Markdown formats.
Audit Metadata