new-skill
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/scaffold.shscript usessedto inject a user-provided description into a template file. It uses|as a delimiter (sed -e "s|{{description}}|$DESC|g"). A user-supplied description containing the|character could break thesedcommand or allow for manipulation of the generatedSKILL.mdfile via sed injection. - [PROMPT_INJECTION]: The skill automates the creation of
SKILL.mdfiles where thedescriptionmetadata field is populated directly from user input without sanitization. This creates a surface for indirect prompt injection if the resulting skill files are later processed by an AI agent that treats the metadata as instructions. - [COMMAND_EXECUTION]: The script executes a local audit tool (
python3 scripts/audit_skills.py). While this is intended behavior for a developer tool, it relies on the integrity of the project's internal scripts and the execution environment.
Audit Metadata