lodestar-setup
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple shell commands via Node.js scripts to analyze the local environment. This includes
gitoperations to determine branch names, commit history, and file churn, as well as package manager commands (npm,pnpm,yarn,bun) to install necessary development dependencies. These actions are limited to standard repository metadata collection and are consistent with the skill's stated purpose of project setup. - [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the
fallowpackage from the official NPM registry. This installation is gated by a user-permission step where the specific command and its consequences are presented for approval before execution. - [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it reads data from the local repository, such as
package.jsoncontents, directory names, and git commit messages. This data is interpolated into the.agents/lodestar/context.mdconfiguration file. The skill mitigates this by using structured JSON state management and enforcing limits on the amount of data projected to the agent's conversation window. - [DYNAMIC_EXECUTION]: The skill uses
spawnSyncwithin its utility scripts to interact with local CLI tools. The commands executed are constructed from a combination of hardcoded safe patterns and discovered repository facts, following the principle of least privilege required for project analysis.
Audit Metadata