README.md

OceansGuard

AI-assisted development guardrails for any repository.

What it solves

• AI-generated changes that accidentally drop existing code
• Lack of global context (only partial files shown)
• Forgetfulness / inconsistent constraints across sessions
• No test / lint guarantees
• Secret leakage (keys/tokens) into commits
• Risky full-rewrite changes

Core commands

init

Create minimal guard files in target repo (idempotent; no overwrite).

python core/aiguard.py init --repo .

pack

Generate AI context pack (diff-first).

python core/aiguard.py pack --repo .

check

Run guard checks + configured project checks and write reports.

python core/aiguard.py check --repo .

run

Shortcut = pack + check.

python core/aiguard.py run --repo . --task "your task"

Strict mode

--strict makes guardrails non-negotiable:

• requires PyYAML
• fails if checks.commands is empty
• fails if contracts/openapi.json is missing/empty

python core/aiguard.py run --repo . --task "CI guard" --strict

Submodule usage (recommended)

In your target repository:

git submodule add https://github.com/OceansCreative/OceansGuard.git tools/OceansGuard
python tools/OceansGuard/core/aiguard.py init --repo .
python tools/OceansGuard/core/aiguard.py run --repo . --task "初回ガード適用"

Outputs

• ai_context_pack.md: single file to paste into AI chat
• ai_test_last.log: raw execution logs
• ai_check_report.json: structured result for CI/PR gating

Git hooks (prevent committing to main)

Install with:

python core/install_hooks.py --repo .