The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill allows users to provide content that is then inserted into generated HTML slide decks. While this represents a surface for prompt injection, it is the intended primary purpose of the skill and is used to process user-supplied copy for their own presentations.
Ingestion points: Slide content placeholders in the generated index.html file.
Boundary markers: Content is placed within standard HTML <section> tags.
Capability inventory: The skill produces a static HTML artifact rendered in the browser.
Sanitization: No explicit sanitization or escaping of user input is specified in the prompt instructions, which is common for user-directed content generation.
[COMMAND_EXECUTION]: The documentation includes command-line snippets for manual verification of the skill registration.
Evidence: curl -s localhost:7456/api/skills | node -e "..." in README.md and SKILL.md.
Context: This command is intended for manual execution by the developer against a local daemon and does not involve automated execution by the agent or skill during runtime.
[SAFE]: The navigation script embedded in the HTML templates handles user interface events (keyboard and scroll) and utilizes local storage to persist the user's position in the deck. It does not perform network operations or access sensitive system files.

replit-deck