The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources which creates a surface for indirect prompt injection.
Ingestion points: The workflow involves reading and extracting data from user-provided Excel, CSV, PDF reports, Word documents, databases, and PowerPoint templates (SKILL.md, Phase 1 & 2).
Boundary markers: No explicit boundary markers or "ignore instructions" delimiters are specified for handling the extracted data.
Capability inventory: The skill uses shell commands for file conversion (soffice, pdftoppm) and writes files to the local system (SKILL.md, Phase 4).
Sanitization: There is no evidence of sanitization, validation, or escaping of the content extracted from external sources before it is interpolated into the deck or processed by the agent.
[COMMAND_EXECUTION]: The skill utilizes shell commands for visual validation loops.
Evidence: The validation phase (SKILL.md, Phase 4) explicitly instructs the execution of soffice --headless --convert-to pdf presentation.pptx and pdftoppm -jpeg -r 150 presentation.pdf slide. These commands process files that are either provided by the user or generated from user data.

pitch-deck