create-skill
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes internal Python scripts to perform lifecycle tasks such as scaffolding new skill directories and validating frontmatter schemas.
- [COMMAND_EXECUTION]: Instructions guide the user and agent in utilizing dynamic context injection via the
!commandsyntax to retrieve real-time system information like git status. - [PROMPT_INJECTION]: The skill processes untrusted user data via
$ARGUMENTSin SKILL.md. - Ingestion points: Phase 0 ($ARGUMENTS)
- Boundary markers: Use of AskUserQuestion to gather requirements and clarify user intent
- Capability inventory: Execution of initialization and packaging scripts; task delegation to linting agents
- Sanitization: Regular expression normalization of skill names in
init_skill.pyto prevent path traversal - [SAFE]: The initialization script (
init_skill.py) normalizes and sanitizes user-provided skill names using regular expressions to mitigate the risk of path traversal or command injection. - [SAFE]: The skill generates script files from predefined local templates rather than executing code from remote or untrusted sources.
Audit Metadata