secret-handling

Preventing AI agents from leaking secrets into git history requires blocking agent access to secret files and preventing writes to .squad directories. This skill enforces no-secret-read policies, promotes placeholder files like env.sample, and adds pre-commit validation to catch exposures before commit.

Code generation agents write credentials to squad logs because they have full repository access and read live secret files. Without explicit prohibition lists and safe alternatives, agents copy secrets into .squad/ files which Scribe auto-commits to remote git history.

Setting up pre-commit validation for secret scanning in repositories involves integrating Scribe pre-commit hooks with regex-based scanning. This skill provides explicit prohibition lists and redaction strategies to validate staged files and block credential leakage before it reaches remote history.

Yes, this secret handling approach works with automation squads using env files by guiding safe configurations across contexts. It prohibits reading live env files, promotes env.template placeholders, and provides escalation guidance for credential exposure remediation.

The best way to manage environment credentials for AI code generation squads is enforcing placeholder environment files and schema documentation references. This skill guides teams on when to ask users for credentials and how to reference schemas instead of reading live secrets.

You should not allow agents to read environment files in repositories when they contain live credentials that could be written to .squad/ files and auto-committed. This skill enforces absolute prohibitions on reading secrets and provides remediation guidance for exposure incidents.