publication-chart-skill
Warn
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions and the 'ensure_publication_tooling.py' script automate the installation of Python packages ('pubfig', 'pubtab') from external registries.
- Evidence: SKILL.md explicitly instructs the agent to 'auto-install it by default' using 'pip' or 'uv' if tools are missing.
- [DYNAMIC_EXECUTION]: The skill uses dynamic module loading and subprocess execution for environment setup and tool management.
- Evidence: 'scripts/ensure_publication_tooling.py' uses 'importlib.import_module()' to probe for packages and 'subprocess.run()' to execute installation commands.
- [COMMAND_EXECUTION]: The skill triggers shell commands to verify tool availability and install dependencies.
- Evidence: SKILL.md contains instructions for executing 'python -c "import pubfig; ..."' and 'pubtab --help' to check the environment.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (Excel files, LaTeX documents, spreadsheets) to generate scientific artifacts and code.
- Ingestion points: User-provided Excel sheets, '.tex' files, and draft plots identified in SKILL.md.
- Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded instructions in external data.
- Capability inventory: The skill can execute shell commands via 'subprocess', perform file writes ('save_figure'), and interact with the Python environment.
- Sanitization: No explicit data sanitization or validation steps are defined for the input files.
Audit Metadata