The Agent Skills Directory

[SAFE]: The skill functions as a benign utility guide for file format conversions. It demonstrates secure coding practices by explicitly using yaml.safe_load() instead of the unsafe yaml.load() method.
[EXTERNAL_DOWNLOADS]: The documentation suggests installing various reputable and widely-used Python packages (e.g., pandas, Pillow, pydub, pyyaml) and system-level tools (e.g., pandoc, ffmpeg, poppler-utils) from official repositories. These are standard dependencies for the tasks described.
[COMMAND_EXECUTION]: Provides examples of shell commands for document conversion tools like pandoc. These are standard use cases for the mentioned utility and do not involve suspicious or unauthorized command execution.
[PROMPT_INJECTION]: The skill identifies an attack surface for indirect prompt injection due to its core function of processing external data files.
Ingestion points: Data is ingested through multiple formats such as CSV, JSON, XML, PDF, HTML, and EPUB using various standard parsers.
Boundary markers: No specific boundary markers or instructions to ignore embedded content are included in the provided code snippets.
Capability inventory: The skill enables file system read/write operations and the execution of external document conversion tools.
Sanitization: The skill promotes the use of yaml.safe_load() for YAML parsing, which is an effective mitigation against deserialization attacks. Other formats use standard, non-executable parsing methods.

file-converter