Team-Ops is experimental (v0.x). Only the latest main is supported; there are no backported fixes for older tags yet.

Team-Ops generates executable code. When it runs, it emits shell scripts under .team-ops/checks/, a git pre-commit hook, and a Stop hook in .claude/settings.json that runs those scripts automatically. These execute on your machine.

Before enabling the gate:

• Read the generated scripts. They are short, plain grep-based checks by design — skim them. Do not enable hooks you haven't read.
• The scripts Team-Ops generates only read your source (grep/file tests) and exit 0/1. If a generated script does anything else — writes files, makes network calls, runs package managers — that is a bug; please report it (see below).
• Hooks run in your shell with your permissions. Treat a cloned Team-Ops repo the way you'd treat any repo whose hooks you're about to enable: review first.

This is inherent to the feature (deterministic enforcement requires running code), not a flaw — but it means review is your responsibility, and the tool is written to keep the generated code small and legible enough to review.

Please do not open a public issue for a security problem.

• Preferred: use GitHub's private vulnerability reporting on this repository (Security tab → "Report a vulnerability").
• Or email the maintainer at saar.upadhyay@gmail.com (replace before publishing).

Include: what the generated artifact does, the input that produced it (repo shape or interview answers), and the impact. We aim to acknowledge within a few days given this is a volunteer, experimental project.

In scope: the generator producing an unsafe or malicious script; the emitters wiring a hook that runs unexpected code; injection through repo contents or interview input that causes the generator to emit something harmful.

Out of scope: vulnerabilities in the AI model or host tool (Claude Code, Codex, Gemini CLI) itself; issues in projects that Team-Ops was run against.