The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands (ls, grep, sed, uv run) to manipulate SVG image files and run formatting checks on the presentation project files.- [DYNAMIC_EXECUTION]: The skill uses a complex shell command to dynamically resolve the path to a Python script (detect_widows.py) located in a hidden plugin cache directory (~/.claude/plugins/cache/...). The script is then executed using uv run. This pattern of executing code from a dynamically calculated path is a risk factor as it relies on the integrity of the file system content at runtime.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided news headlines and quotes, creating a surface where untrusted data could potentially influence agent behavior during processing.
Ingestion points: The skill reads news headline and quote content from data/headlines.json to populate presentation cards.
Boundary markers: There are no specific delimiters or instructions provided to the agent to treat the headline/quote text as untrusted data or to ignore any instructions embedded within that text.
Capability inventory: The skill environment includes file writing capabilities and the ability to execute shell commands and Python scripts (uv run, sed, grep).
Sanitization: The instructions do not specify any validation or sanitization steps for the headline or quote text before it is written to the data file or processed by the rendering engine.

headline-card