conversation-json-to-md
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided JSON chat exports. Malicious instructions embedded in the chat history could be interpreted by the agent when it performs the mandatory second-pass verification and formatting pass. \n
- Ingestion points: The skill reads input JSON files containing user and assistant chat history via the scripts/convert_conversations.py script. \n
- Boundary markers: The conversion script uses Markdown headings (## and ###) to separate questions and answers, providing structural delimiters. \n
- Capability inventory: The skill allows the agent to read and write files in the workspace and execute a bundled Python script. \n
- Sanitization: Filenames are sanitized for illegal characters and length, but the chat content itself is preserved to maintain Markdown integrity as intended.\n- [COMMAND_EXECUTION]: The skill executes a bundled local script scripts/convert_conversations.py to handle the JSON parsing and file generation. The command parameters are defined by the user for input and output locations.\n- [SAFE]: No high-severity issues such as network exfiltration, hardcoded credentials, or obfuscated code were found. The skill operations are local and limited to the stated purpose of document conversion.
Audit Metadata