The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/convert_conversations.py to perform the conversion. The script uses standard Python libraries (such as json, re, and pathlib) and performs local file system operations, including reading the input JSON and writing or unlinking Markdown files in the output directory.
[SAFE]: No network activity, hardcoded credentials, or obfuscated code were detected. The script performs deterministic structural transformations and includes sanitization logic for generated filenames to prevent illegal character issues.
[SAFE]: The skill identifies a surface for indirect prompt injection as it processes untrusted conversation data into Markdown output.
Ingestion points: The scripts/convert_conversations.py script reads user-provided JSON files via argparse and json.load.
Boundary markers: Absent. The script formats content using Markdown headers (##, ###) but does not include explicit safety delimiters or warnings within the output files.
Capability inventory: Local file system write (write_text) and specific file deletion (unlink for *.md files) via the pathlib.Path module.
Sanitization: Filenames are sanitized for illegal characters and length, while message content is preserved as Markdown without escaping. This is consistent with the skill's primary purpose as a structural formatting utility.

conversation-json-to-md