skill-sync
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the execution of a local utility named
skill-syncand various agent command-line interfaces, includingclaude,codex, anddevin. It specifically supports the--yoloflag for Devin, which allows for autonomous execution with reduced human oversight. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from a user-provided source path to scaffold new skills. This creates an attack surface where instructions embedded within the source files could influence the agent's behavior during the build process.
- Ingestion points: Content from the directory specified in
<source-path>(SKILL.md). - Boundary markers: None explicitly defined in the instructions for agent interpolation.
- Capability inventory: File system writes (rsync mirroring, new file creation), command execution (invoking agent CLIs), and network operations (git push, GitHub PR creation).
- Sanitization: No sanitization or validation of the source material is mentioned prior to processing by the target agent.
- [EXTERNAL_DOWNLOADS]: The installation process involves downloading the skill via the NPM registry (
npx skills@latest) and executing a local installation script (install.zsh) to set up the companion utility.
Audit Metadata