harness-init
Pass
Audited by Gen Agent Trust Hub on Jul 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill recommends installing related agent skills (e.g.,
RayChang/agent-skills@kb-wiki) as a post-scaffolding step in its report format. These resources are authored by the same vendor and are intended for project management. - [COMMAND_EXECUTION]: The generated templates include standard development commands for Git operations (worktree creation, branch management) and package managers (npm, pnpm, bun, yarn). These commands are explicitly defined within a gated development pipeline in the
workflow.mdtemplate, which requires human approval before execution. - [PROMPT_INJECTION]: The skill populates project templates by extracting metadata from
package.jsonand local Git configurations. While this creates a potential surface for indirect prompt injection if these local files are maliciously crafted, the skill's operational context is local project scaffolding with high user visibility. - [SAFE]: The skill implements a strict additive-only rule, preventing any accidental modification or loss of existing project data during the initialization process.
Audit Metadata