What problem does it solve? Running a coding agent (Claude Code, Codex) against real credentials risks leaking API keys into transcripts, files, or commits. This Skill explains how to use dotenvx to inject, encrypt, and redact .env secrets, and — critically — what its redaction does and does not protect against. ## Core Features & Use Cases - Output redaction: Run agents via dotenvx run --redact -- <agent> so literal secret values in stdout/stderr are replaced with [REDACTED], with honest coverage of its exact-substring-matching limits. - Encryption and key handling: Encrypt .env files with dotenvx encrypt, manage .env.keys and DOTENV_PRIVATE_KEY per environment, and inject keys through CI or deployment vendors. - Vault references and guards: Resolve op:// (1Password) and bw:// (Bitwarden) references, enforce 1Password service-account vault access rules, and install dotenvx gitignore/precommit guards. - Use Case: Before letting an agent run database migrations with production credentials, wrap it in dotenvx run --redact --overload -- claude -p '...', keep .env encrypted, and store the private key in 1Password. ## Quick Start Ask the agent to run a command through dotenvx with redaction enabled, for example: wrap my next dev server in dotenvx run --redact --overload so my .env secrets are injected but never echoed back.