obsidian-vault-add
Warn
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches metadata such as descriptions, biographies, and titles from external sources including Wikipedia, OpenLibrary, iTunes, TVMaze, and general Web Search, and inserts this content directly into Obsidian notes.
- Ingestion points: Step 1 (Web Search for metadata) and Step 7 (API calls to Wikipedia, OpenLibrary, etc.).
- Boundary markers: No delimiters or instructions to ignore embedded commands are used when inserting fetched text into notes.
- Capability inventory: The agent has access to Bash, python3, Edit, Write, and specialized CLI tools (obsidian, qmd).
- Sanitization: There is no evidence of sanitization, escaping, or validation of the content retrieved from the web before it is written to the vault.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to update local Python scripts (e.g., 99 Meta/scripts/download-book-covers.py) by appending new entries derived from external metadata.
- Evidence: Step 7 explicitly directs the agent to 'add the new entry to the appropriate download script so future re-runs include it'.
- Risk: Appending untrusted data (like book titles or author names) to an executable script creates a code injection vector. If the fetched metadata contains malicious Python code, it could be executed during subsequent runs of the download scripts.
Audit Metadata