defuddle
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a local Python script (
claude-obsidian.py) located at a fixed installation path to perform its functions. It includes defensive coding practices, such as instructing the agent to never interpolate URLs into shell strings and to use single argv elements to prevent command injection. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, specifically user-provided URLs and local HTML/Markdown files. It mitigates the risk of malicious instructions hidden in this data by enforcing an inert 'dry run' planning phase, requiring manual approval of transaction hashes, and explicitly rejecting sensitive query parameters or credentials from inputs.
Audit Metadata