The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes marp-cli and mermaid-cli to generate presentations and diagrams. It also uses brew install to manage these dependencies locally.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes Markdown (.md) and Mermaid (.mmd) files, which serves as a potential surface for instructions embedded in data to influence the agent's operations. The analysis identified the following evidence chain:
Ingestion points: Reading existing .md files in docs/ and .mmd files (SKILL.md).
Boundary markers: None explicitly defined in instructions.
Capability inventory: Execution of marp-cli, mermaid-cli, and brew commands via subprocess (SKILL.md).
Sanitization: Standard tool-based rendering without explicit agent-side filtering.

pavlo-marp-presentation