The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The skill specifies installation of the markitdown package and its dependencies. Since Microsoft is a trusted organization, this dependency is classified as low risk per the [TRUST-SCOPE-RULE].
[PROMPT_INJECTION] (LOW): This skill presents an indirect prompt injection surface (Category 8) because it extracts content from untrusted external documents (PDF, DOCX, etc.) to be processed by an LLM.
Ingestion points: File conversion via md.convert() in scripts/batch_convert.py and SKILL.md examples.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are implemented in the conversion logic.
Capability inventory: File system read/write access and network access for Azure and OpenAI integrations.
Sanitization: No explicit sanitization or escaping of extracted document text is performed before it is provided to the agent.
[COMMAND_EXECUTION] (SAFE): The provided scripts/batch_convert.py utility follows best practices for command-line argument parsing and file system interactions; no arbitrary command injection vulnerabilities were found.
[CREDENTIALS_UNSAFE] (SAFE): API keys for Azure and OpenAI are appropriately handled via environment variables and standard CLI arguments, with no hardcoded secrets detected in the source code.

markitdown