claudian

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill facilitates full shell access (bash mode) and file system operations (read/write/edit) within the local environment of the Obsidian vault. These capabilities are intended for the 'agentic collaborator' workflow but grant significant control over the local machine.
  • [PROMPT_INJECTION]: The skill has a high surface area for indirect prompt injection attacks.
  • Ingestion points: Directly reads and processes files from the Obsidian vault (SKILL.md).
  • Boundary markers: Not explicitly defined in the configuration, though it allows for a 'Custom system prompt' which may be used to set constraints.
  • Capability inventory: Includes bash execution, file reading, file writing, and search operations (grep, glob) across the vault.
  • Sanitization: Provides safety features such as a 'Command blocklist' and 'Allowed export paths' to limit the scope of potential exploitation.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 02:49 PM