process-inbox
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from user notes and external configuration files.\n
- Ingestion points: Reads files from the Obsidian vault (Inbox notes) and an 'AGENTS.md' configuration file to determine processing logic.\n
- Boundary markers: The instructions do not specify delimiters or system-level warnings to distinguish between note content and agent instructions.\n
- Capability inventory: Execution of file move and delete commands via the Obsidian CLI.\n
- Sanitization: No explicit sanitization or filtering of note content is performed before summarization or classification; however, the workflow mandates explicit user approval for every proposed action.\n- [COMMAND_EXECUTION]: The skill uses a command-line interface to modify the local file system.\n
- Evidence: Instructions in 'SKILL.md' (Step 3) direct the agent to use the Obsidian CLI for executing approved file movements and deletions.
Audit Metadata