The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from external sources to generate automated outputs. \n- Ingestion points: The skill explicitly instructs the agent to extract values directly from provided Excel and CSV source files (SKILL.md). \n- Boundary markers: There are no specified delimiters or instructions to treat data from these files as purely passive content or to ignore potential embedded instructions. \n- Capability inventory: The skill utilizes the 'pptx' tool to create PowerPoint objects, including charts and tables, which could be influenced by malicious data content (SKILL.md). \n- Sanitization: The instructions lack requirements for validating, escaping, or sanitizing the content of the external data files before they are processed and interpolated into the final presentation.

competitive-analysis