The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Python scripts (scripts/presentation_check.py and scripts/html_deck_bridge.py) to synchronize deck data and perform mechanical quality checks. These scripts are implemented using only the Python standard library and include internal security controls such as path traversal validation to ensure operations remain within the designated project root.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data provided via input source files and local HTML slide files, creating a potential attack surface that is mitigated by specific design choices.
Ingestion points: Reads text content from brief.json (sources field) and statically parses HTML content from the slides/ directory.
Boundary markers: The instructions in SKILL.md explicitly state that commands found within source materials must be treated only as evidence and should not be executed.
Capability inventory: The skill can read and write files within the project scope and execute its own local validation scripts.
Sanitization: The accompanying Python scripts implement rigorous path validation (is_relative_to), explicitly block the loading of external network resources in HTML processing, and use a hardened JSON parser to prevent structure-based injection attacks.

presentation-ja