setup
Warn
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill's detector logic in 'scripts/detector.ts' executes shell commands retrieved from plugin manifests ('manifest.json') to verify the status of CLI dependencies. This allows for the execution of arbitrary commands defined in metadata by any installed plugin.
- [REMOTE_CODE_EXECUTION]: The application implements an update feature in 'playground/src/lib/app-update-install.ts' that downloads disk images from bopen.ai, mounts them using system tools, and executes a generated shell script to replace the running application bundle.
- [EXTERNAL_DOWNLOADS]: The skill performs network requests to bopen.ai to fetch marketplace data, download software updates, and retrieve UI sound assets. It also references GitHub repositories for installing third-party skills.
- [DATA_EXFILTRATION]: The skill accesses local usage logs ('~/.claude/core/skill-activity.jsonl') and audits the presence of environment variable keys to report on harness health. While it explicitly avoids displaying secret values, this exposure provides the agent with metadata about system configuration and usage patterns.
Audit Metadata