managing-obsidian-tasks

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill interacts with the local environment by executing the obsidian or obsidian-cli binary using subprocess.run. This is the primary mechanism for vault interaction, including reading, creating, and updating notes.
  • [DYNAMIC_EXECUTION]: In scripts/obsidian_adapter.py, the skill dynamically generates JavaScript code blocks to be executed within the Obsidian application context via the CLI's eval command. The implementation uses Base64 encoding for data payloads to prevent command injection into the JavaScript context.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from user-controlled Obsidian vaults. Maliciously crafted task notes could potentially contain instructions aimed at misleading the agent. The skill mitigates this by enforcing a strict schema (task contract) and requiring confirmation for mutations.
  • Ingestion points: Vault search results and note content are read in scripts/obsidian_adapter.py and processed in SKILL.md (Step 2: Inspect task context).
  • Boundary markers: The instructions mandate adherence to references/task-contract.md and use specific views for status transitions.
  • Capability inventory: The agent can read/write files and execute shell commands through the Obsidian CLI.
  • Sanitization: Content is transported using Base64 to prevent shell injection, and the agent is instructed to avoid interpolating markdown into shell arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 04:05 PM