lwy-project-agent-writer
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local, bundled Node.js script (
scripts/cli.cjs) to render markdown templates for new agents. This execution is performed only after the user confirms the proposed design and is restricted to writing files within the project's.agents/agents/directory. - [PROMPT_INJECTION]: The skill analyzes external project files (such as
package.jsonand directory structures) to propose contextual agent designs. While this ingestion of potentially untrusted project data creates a surface for indirect prompt injection, the risk is mitigated by a requiredAskUserQuestionconfirmation phase where the user reviews the design before generation. - [SAFE]: The skill follows security best practices by declaring its side effects, enforcing strict boundaries for file operations, and maintaining a human-in-the-loop requirement for all modifications.
Audit Metadata