obsidian-literature-base
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from literature metadata and abstracts to generate AI-assisted topics.
- Ingestion points: Metadata JSON files (metadata.json), MinerU-generated Markdown documents, and Zotero reading records found within the Obsidian vault.
- Boundary markers: The skill uses ownership markers such as the generatedBy YAML property and HTML comments ( and ) to delimit auto-generated sections.
- Capability inventory: The script performs file system operations including directory creation and file writes/overwrites within the vault root using pathlib and os.replace.
- Sanitization: The implementation uses yaml.safe_load for parsing and includes dedicated sanitization functions (_safe_wiki_stem, _obsidian_tag) to clean filenames and tags using regular expressions and Unicode normalization.
Audit Metadata