The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches content from external URLs provided in the AGENDA.md file to generate source summaries for slides.
[COMMAND_EXECUTION]: The skill performs shell command checks for environment tools (node, marp) and executes the marp-cli tool to convert markdown into a final HTML presentation.
[PROMPT_INJECTION]: The skill processes untrusted web content, which presents a surface for indirect prompt injection attacks.
Ingestion points: External research content is fetched from URLs and stored in docs/sources/ (detailed in SKILL.md and SOURCES.md).
Boundary markers: The skill includes clear instructions to the agent to treat fetched content as untrusted data and explicitly ignore any directives found within it.
Capability inventory: The skill uses the marp command-line tool to process the final presentation markdown.
Sanitization: The SOURCES.md file contains a dedicated security section with patterns to detect and skip malicious instructions in fetched content, such as "ignore previous instructions" or role-play attempts.
Note on Static Detection: A static analysis flag for instruction override was found to be a false positive; the identified text is part of the skill's own security filtering logic.

generate-slides