lov-personal-vocabulary
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data from dictation app dictionary files and API responses to maintain a central vocabulary list.
- Ingestion points: The skill reads
vocabulary.json, OpenLessdictionary.jsonfiles, and JSON responses from the Typeless API via theapp-adaptersmodule. - Boundary markers: The
SKILL.mdworkflow includes a mandatory step for the agent to 'confirm before execute' (先确认再执行) for synchronization plans, providing a human-in-the-loop safety gate, though specific input delimiters within the processed JSON fields are not explicitly defined. - Capability inventory: The skill possesses file read and write capabilities for local JSON data storage and configuration management, implemented via the
vocab_cli.pyandprofile_store.pyscripts. - Sanitization: Data processing is performed using standard Python JSON libraries with specific field mapping and structure validation; the skill does not execute code or shell commands derived from the content of the vocabularies.
Audit Metadata