aep-e2e-skill-scaffolding
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill generates multiple shell scripts, including
seed.sh,preflight.sh, andderive-verification-recipe.sh, by rendering local templates with project-specific data. It useschmod +xto grant execution permissions to these generated scripts, which is standard behavior for a scaffolding utility. - [COMMAND_EXECUTION]: The skill utilizes shell commands to manage the project's file structure. This includes detecting the current state of the filesystem, migrating legacy directories using
git mv, and creating symbolic links to ensure compatibility across different agent platforms. - [EXTERNAL_DOWNLOADS]: The skill's documentation references several third-party automation tools such as
agent-browser(from Vercel Labs) andwebwright(from Microsoft). These references are informative and point to official repositories of well-known and trusted organizations. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from project files like
package.jsonto populate its templates. To mitigate the risk of malicious configuration values, the skill explicitly proposes these values to the user and requires manual confirmation or adjustment before rendering the final files.
Audit Metadata