Our dev agents (OpenCode scripts and plugins) read tokens and passwords from environment variables, which any Windows process can read. What is the best-practice, easy-to-use secret manager for scripts, plugins, and agents?
This report compares 8 vault and CLI interfaces documented across 5 official-documentation hosts (1Password, Doppler, Infisical, HashiCorp, Microsoft; 294 documentation pages scanned), on how each stores secrets, authenticates unattended automation, and — critically — in what form plaintext reaches the running process.
Operational complexity rises left to right. Colour shows how plaintext reaches the process at execution time — none of these makes a delivered secret unreadable to an equally privileged process, so pair any of them with a low-privilege agent account and brokered tools.
Hover or tap an option to see its storage, unattended auth, and script integration here.
.env; split identities per tool, project, and environment, and make production access read-only or approval-gated.deploy(target), query_database(read_only_query) — never get_secret(name); trusted code uses the secret internally and returns sanitised results.| Option | Script integration | Unattended auth | Plaintext reaches process as | Doc pages | Source |
|---|
Comparison of 8 secret-management products and interfaces, one row each, extracted from official documentation on 5 hosts (294 pages scanned) during this research. “Doc pages” counts documentation pages found in the extraction, a rough proxy for surface area and operational complexity, not quality. Full storage and exposure descriptions are shortened in the lane board; exact wording appears in the detail panel and table. Each row links its documentation source.