registry-publish
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions specify the use of shell commands such as
npm pack,npm publish, andnpm install. These commands are used to prepare the package, publish to the registry, and verify the artifact. - [DYNAMIC_EXECUTION]: In Step 6, the skill performs post-publish verification by executing code from the newly installed package using
node -e "import('<name>/<entry>').then(m => console.log(m))". This involves executing code at runtime based on the package's entry points. - [INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface where it ingests and executes data (the package code) from an external registry. During the verification step, if the package content were to be malicious, it could attempt to influence the agent's behavior.
- Ingestion points: Package entry points defined in
package.json(Step 6). - Boundary markers: None explicitly defined for the verification execution.
- Capability inventory: Uses
npm installandnode -e(Step 6). - Sanitization: Not present; the code is imported and executed as-is for verification purposes.
Audit Metadata