Project Iron Bootstrap
Inject the reusable Project Iron governance core from
git@github.com:sy007-spec/meta-injector.git into another git repository. The
backend stages working copies of meta-injector and your target repo, runs
scripts/bootstrap_project.py, then commits and pushes the result — so the
target project immediately has an AI-agent-ready development baseline.
What meta-injector does
scripts/bootstrap_project.py <target> initializes git in the target if needed, adds meta-injector as a git submodule at vendor/project-iron-core, then runs the embedded project-iron-skill pipeline (init → sync → doctor).
init scaffolds the governance surfaces: PROJECT_IRON_RULES_SPEC.md, an append-only PROMPT_INPUT_LOG.md, AGENTS.md (Codex), CLAUDE.md, Cursor + VSCode rule files, a machine-readable PROJECT_LLM_REQUIREMENTS.json, a unified ops.py entrypoint, DOMAIN_IRON_RULES.md + skills/README.md starting points, and a dated LLM_OUTPUTS/ archive.
sync re-applies the template governance files on every run but never overwrites ops.py, PROJECT_LLM_REQUIREMENTS.json, or PROMPT_INPUT_LOG.md once they hold real project state. doctor verifies the binding and fails on drift.
Note: the committed .gitmodules records meta-injector's URL as the server sees it. If that repo is private, anyone cloning your target later needs read access to it (or make meta-injector public).
This tool runs server-side git + python and pushes commits. It requires a bo-admin session and (in production) a request from an allowlisted IP.