The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a defensive layer against prompt injection attacks originating from external websites. The SOURCES.md file instructs the agent to scan fetched content for malicious directives (e.g., 'ignore previous instructions', 'act as') and skip processing if detected.\n- [EXTERNAL_DOWNLOADS]: The skill fetches data from user-provided URLs to generate slide summaries. This process is restricted by a security policy in SOURCES.md that treats all fetched content as untrusted and limits extraction to factual data.\n- [COMMAND_EXECUTION]: The skill executes the marp command-line utility to generate presentation outputs. This is the core functionality and is restricted to the local project environment.

generate-slides