bring-your-own-agent
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands for discovery and verification (e.g.,
curl,grep,python manage.py). Evidence:discovery.mdandcredentials.mdcontain commands to probe app routes and test credentials. These are used as part of the intended development workflow for the owner to verify their own app's API. - [EXTERNAL_DOWNLOADS]: The skill instructs the installation of standard Node.js packages. Evidence:
agent-access/package.jsonincludes@modelcontextprotocol/sdk,dotenv, andzod. These are well-known, industry-standard libraries for building MCP servers and managing configuration. - [DATA_EXFILTRATION]: The skill includes explicit rules to prevent data exfiltration. Evidence:
SKILL.mdandserver.mdmandate that secrets never leave a tool, return only specific fields (no spreads), and log only tool arguments, never credentials. - [CREDENTIALS_UNSAFE]: The skill follows secure credential management practices. Evidence:
server.mdgenerates a.envfile that is explicitly gitignored in the generated folder's own.gitignore, ensuring secrets are not committed to the repository. - [SAFE]: The skill is authored by 'brainit-consulting' and all instructions are oriented toward building local, user-controlled infrastructure that talks to the user's existing applications. It includes specific 'Step 0' safety checks to recommend better in-app integration paths where available.
Audit Metadata