translate-skills
Pass
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands like
findandpwdto scan forSKILL.mdfiles across multiple global and local directories, including~/.claude/skills,~/.cursor/skills, and~/.trae/skills. - [PROMPT_INJECTION]: As an indirect injection surface, the skill reads and processes the content (frontmatter) of external
SKILL.mdfiles. This metadata could potentially contain instructions designed to influence the agent's behavior during the translation and writing process. - Ingestion points: Reads
SKILL.mdfiles from local and global directories via file system tools. - Boundary markers: Not explicitly defined for the content being read.
- Capability inventory: Uses file-edit tools to modify contents and shell tools to find files.
- Sanitization: No specific sanitization or validation of the input fields (
name,description) is mentioned before processing.
Audit Metadata