llm-wiki
Warn
Audited by Gen Agent Trust Hub on May 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for establishing a persistent background process using a systemd user service ('obsidian-wiki-sync.service') to handle continuous synchronization of wiki files.
- [COMMAND_EXECUTION]: The skill utilizes the 'sudo loginctl enable-linger $USER' command to ensure that user-level background services continue to run after the user logs out.
- [EXTERNAL_DOWNLOADS]: The documentation directs the installation of the third-party package 'obsidian-headless' from the public NPM registry to enable headless synchronization.
- [EXTERNAL_DOWNLOADS]: The 'Ingest' operation uses the 'web_extract' tool to fetch content from external URLs and remote PDF files for incorporation into the knowledge base.
- [CREDENTIALS_UNSAFE]: The configuration workflow for synchronization involves a command ('ob login') that requires the input of sensitive user credentials, such as email and password, as plaintext command-line arguments.
- [REMOTE_CODE_EXECUTION]: The 'Lint' feature is designed to utilize the 'execute_code' tool to run dynamic Python logic for scanning files, validating metadata, and auditing the wiki structure.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data from external sources ('raw/' directory) and uses it to update wiki pages. This ingestion occurs without explicit boundary markers or sanitization to prevent malicious instructions within the source material from influencing the agent during subsequent read and synthesis operations.
Audit Metadata