The Agent Skills Directory

[COMMAND_EXECUTION]: Several components within the toolset use subprocess.run to coordinate multi-stage tasks. For example, scripts/apply_template.py invokes the svg_quality_checker.py script to validate templates during installation, and maintenance scripts in scripts/docs/ use shell execution to run test suites against the toolset's internal utilities.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and restructure data from external sources, creating a surface for indirect prompt injection. Ingestion points include utilities in scripts/source_to_md/ that extract content from PDF, DOCX, XLSX, and HTML files. The system relies on structured Markdown headers and block-level metadata but does not implement specific adversarial delimiters for untrusted content. The skill possesses capabilities for network interaction, file system modification, and subprocess execution. While the skill validates SVG structure, it does not explicitly sanitize ingested text for prompt injection patterns.\n- [EXTERNAL_DOWNLOADS]: The skill communicates with various external APIs and web providers for resource acquisition. It integrates with multiple established AI image generation backends (e.g., OpenAI, Google Gemini, Stability AI) and fetches images from well-known providers such as Pexels and Pixabay via scripts/image_search.py.

ppt-master