The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill explicitly instructs the agent to fetch and read external files, URLs, or references provided by the user to populate the slide deck content (Phase 1, Question 3).
Ingestion points: The agent is directed to "read/fetch them thoroughly" if the user provides external data during the interview phase.
Boundary markers: No specific boundary markers or instructions to ignore embedded prompts are provided for the fetched content.
Capability inventory: The skill utilizes shell execution via npx and python3, and performs file writing/modification operations in SKILL.md.
Sanitization: While the scripts/marp-mobile.py script uses html.escape for breadcrumb and title fields, it enables html: True in the MarkdownIt renderer, which may allow HTML tags from the processed data to be rendered in the mobile export.
[COMMAND_EXECUTION]: The skill uses multiple subprocess calls to perform its tasks.
Evidence: Usage of npx @marp-team/marp-cli@latest to generate HTML/PPTX exports and python3 scripts/marp-mobile.py to generate the mobile view.
[DYNAMIC_EXECUTION]: The skill executes dynamically defined scripts to handle file manipulation and post-processing.
Evidence: SKILL.md uses a python3 heredoc (<<'EOF') to inject JavaScript into exported HTML files and a python3 -c one-liner to perform object-oriented manipulation of PPTX files using the python-pptx library.
[EXTERNAL_DOWNLOADS]: The skill relies on external packages and tools downloaded from official registries.
Evidence: Downloads the marp-cli package via npx and requests the installation of markdown-it-py and python-pptx via pip. These originate from well-known registries (NPM and PyPI).

create-marp-deck