Thanks for your interest - AIGX is meant to be a shared standard, and it gets better with more eyes, more languages, and independent replication.
- Spec feedback. Found an ambiguity in SPEC.md? Open an issue. Concrete wording proposals (PRs against the spec) are very welcome.
- New worked examples. A genome for a Python / Go / Rust / monorepo project is hugely valuable. Add it
under
examples/<name>/with a short README explaining the codebase it describes. - Tooling. Exporters (
aigx → AGENTS.md / CLAUDE.md / .cursor/rules), a linter (aigx-lint: validate parity, dangling<check>ids, missing entries), editor integrations. - Replication. The strongest contribution of all: reproduce the benchmark on your own codebase / model and report what you find - confirming or challenging the result. Open an issue with your setup and numbers.
- Docs. Clarity fixes, typos, better diagrams.
AIGX is an empirical project. The bar for adding complexity to the format is high, because the benchmark repeatedly showed that embellishments don't help (see principles). If you propose a new field, mechanism, or structure:
- State the failure mode it fixes.
- Prefer the simplest version.
- Ideally, bring evidence (even a small A/B on your own repo) that it changes agent behavior. Remember
L4:
n=30is noise - power it up before trusting it.
"It feels more complete" is not, by itself, a reason - completeness lost to terseness every time we tested.
- Open an issue to discuss anything non-trivial before a big PR.
- Fork, branch, commit with clear messages.
- Keep PRs focused (one idea per PR).
- For spec changes, update
SPEC.mdand bump the version note if it's normative. - Be excellent to each other - see CODE_OF_CONDUCT.md.
By contributing, you agree your contributions are licensed under the MIT License.