prompt-craft
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The README provides installation instructions using
npxto fetch the skill from the vendor's repository (m1nga/skills), which is a standard procedure for the supported platform. - [COMMAND_EXECUTION]: The skill maintains a local configuration directory (
~/.prompt-craft/) to store ASR correction dictionaries and user context. It includes logic to update these files based on user confirmation, which is a benign feature for personalizing tool performance. - [SAFE]: The skill ingests untrusted user input to generate prompts for other agents. It employs structured XML tags and mandatory 'execution discipline' blocks to create clear boundaries and manage downstream agent behavior. Ingestion: user input and context files; Boundaries: XML formatting and explicit disciplinary blocks; Capability: text transformation; Sanitization: filler word removal and ASR normalization.
Audit Metadata