The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts using shell commands to validate figure specifications and audit render results.
Evidence: SKILL.md calls python3 academic-figure-designer/scripts/validate_figure_spec.py.
Evidence: references/render-audit.md and SKILL.md mention running scripts/validate_render_audit.py and clean_image_metadata.py.
The skill contains an explicit security instruction: "Never interpolate a prompt or user-controlled label into a shell command."
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from potentially untrusted sources such as paper URLs, PDFs, and external repositories to generate figure plans.
Ingestion Points: SKILL.md defines routes for "Complete Manuscript (Markdown / LaTeX / PDF / URL)", "Repository path or repository URL", and "Draft Notes / Outline".
Boundary Markers: The skill utilizes strictly defined schemas (FigurePlan v1, FigureSpec v1, RenderAudit v2) to structure processed data before execution.
Capability Inventory: The agent has the capability to execute Python scripts for validation and call native image generation tools.
Sanitization: The workflow includes a mandatory step to run clean_image_metadata.py to strip metadata from generated images and uses structured arguments for CLI calls.
[DYNAMIC_EXECUTION]: The skill includes a standalone Python script (scripts/validate_render_audit.py) intended to be executed during the workflow to verify artifact integrity.
Evidence: The script performs SHA256 hashing and JSON schema validation on generated figure artifacts.
The script uses standard libraries and is provided as a local utility for the agent to ensure scientific and structural accuracy.

academic-figure-workflow