The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted content such as research papers, PDFs, and web pages.
Ingestion points: The agent reads tagged files and user instructions to generate slide content as defined in SKILL.md.
Boundary markers: The skill contains explicit instructions in SKILL.md to treat text found in papers and sources as material only and never as instructions, providing a conceptual boundary.
Capability inventory: The skill utilizes an external PPTX authoring tool and executes a bundled Python script, scripts/check-deck.py, for structural validation.
Sanitization: No technical sanitization of source text is specified beyond the instructional boundary marker.
[COMMAND_EXECUTION]: The skill utilizes local shell commands to perform structural checks and execute tests.
Evidence: SKILL.md instructs the agent to run python3 scripts/check-deck.py after the presentation is saved to verify its structure.
Evidence: The file scripts/tests/test_check_deck.py uses subprocess.run to invoke the validation script during unit testing.
[DYNAMIC_EXECUTION]: The test suite employs dynamic module loading to verify the local validation script.
Evidence: The file scripts/tests/test_check_deck.py uses importlib.util to load and execute scripts/check-deck.py via exec_module. This is a standard procedure for testing local components using a static relative path.

rk-academic-slides